@nodaro/sdk 2.63.0 → 2.65.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (83) hide show
  1. package/dist/{chunk-R63TTR35.cjs → chunk-3CB4WYEY.cjs} +2 -2
  2. package/dist/{chunk-R63TTR35.cjs.map → chunk-3CB4WYEY.cjs.map} +1 -1
  3. package/dist/{chunk-72ZJ72WF.js → chunk-3HL3IHVE.js} +20 -5
  4. package/dist/chunk-3HL3IHVE.js.map +1 -0
  5. package/dist/{chunk-3OARKXWT.cjs → chunk-4DCXJ77J.cjs} +6 -4
  6. package/dist/chunk-4DCXJ77J.cjs.map +1 -0
  7. package/dist/{chunk-T7N7DFQJ.js → chunk-CRIQLLR2.js} +2 -2
  8. package/dist/{chunk-T7N7DFQJ.js.map → chunk-CRIQLLR2.js.map} +1 -1
  9. package/dist/{chunk-ZAATTX5W.js → chunk-DHJRS65Q.js} +5 -3
  10. package/dist/chunk-DHJRS65Q.js.map +1 -0
  11. package/dist/{chunk-K2VWXAGK.js → chunk-HAS6PMQN.js} +6 -4
  12. package/dist/chunk-HAS6PMQN.js.map +1 -0
  13. package/dist/{chunk-GHVCV4S5.cjs → chunk-HVF5DRGQ.cjs} +2 -2
  14. package/dist/{chunk-GHVCV4S5.cjs.map → chunk-HVF5DRGQ.cjs.map} +1 -1
  15. package/dist/{chunk-5LI4CVJB.cjs → chunk-HW3CSUVC.cjs} +2 -2
  16. package/dist/chunk-HW3CSUVC.cjs.map +1 -0
  17. package/dist/{chunk-3QFTP5P5.js → chunk-OHC4JIUP.js} +10 -2
  18. package/dist/chunk-OHC4JIUP.js.map +1 -0
  19. package/dist/{chunk-H3HUZT3T.js → chunk-OVFH7L2T.js} +2 -2
  20. package/dist/{chunk-H3HUZT3T.js.map → chunk-OVFH7L2T.js.map} +1 -1
  21. package/dist/{chunk-GA4LZZKQ.cjs → chunk-PGT4BXR2.cjs} +10 -2
  22. package/dist/chunk-PGT4BXR2.cjs.map +1 -0
  23. package/dist/{chunk-PTVUYLJO.cjs → chunk-QTO2U6JZ.cjs} +5 -3
  24. package/dist/chunk-QTO2U6JZ.cjs.map +1 -0
  25. package/dist/{chunk-POUMH7QE.js → chunk-QUYZCZ33.js} +10 -10
  26. package/dist/{chunk-POUMH7QE.js.map → chunk-QUYZCZ33.js.map} +1 -1
  27. package/dist/{chunk-CAQ2BKEP.js → chunk-S4BERJXU.js} +3 -3
  28. package/dist/{chunk-CAQ2BKEP.js.map → chunk-S4BERJXU.js.map} +1 -1
  29. package/dist/{chunk-A42R7UT7.js → chunk-SVWEWBXI.js} +2 -2
  30. package/dist/chunk-SVWEWBXI.js.map +1 -0
  31. package/dist/{chunk-JK3YVG6A.cjs → chunk-UIM5SLPP.cjs} +20 -5
  32. package/dist/chunk-UIM5SLPP.cjs.map +1 -0
  33. package/dist/{chunk-7IZ4ZMI5.cjs → chunk-XZYJBJ2K.cjs} +18 -18
  34. package/dist/{chunk-7IZ4ZMI5.cjs.map → chunk-XZYJBJ2K.cjs.map} +1 -1
  35. package/dist/{chunk-VU23J7US.cjs → chunk-ZVG5ODVY.cjs} +3 -3
  36. package/dist/{chunk-VU23J7US.cjs.map → chunk-ZVG5ODVY.cjs.map} +1 -1
  37. package/dist/entries/audio.cjs +2 -2
  38. package/dist/entries/audio.d.cts +10 -0
  39. package/dist/entries/audio.d.ts +10 -0
  40. package/dist/entries/audio.js +1 -1
  41. package/dist/entries/cast-swap.cjs +2 -2
  42. package/dist/entries/cast-swap.d.cts +52 -17
  43. package/dist/entries/cast-swap.d.ts +52 -17
  44. package/dist/entries/cast-swap.js +1 -1
  45. package/dist/entries/client.cjs +14 -14
  46. package/dist/entries/client.js +9 -9
  47. package/dist/entries/content-rights.cjs +2 -2
  48. package/dist/entries/content-rights.d.cts +4 -2
  49. package/dist/entries/content-rights.d.ts +4 -2
  50. package/dist/entries/content-rights.js +1 -1
  51. package/dist/entries/core.cjs +5 -5
  52. package/dist/entries/core.js +1 -1
  53. package/dist/entries/jobs.cjs +2 -2
  54. package/dist/entries/jobs.d.cts +24 -8
  55. package/dist/entries/jobs.d.ts +24 -8
  56. package/dist/entries/jobs.js +1 -1
  57. package/dist/entries/media.cjs +2 -2
  58. package/dist/entries/media.d.cts +78 -6
  59. package/dist/entries/media.d.ts +78 -6
  60. package/dist/entries/media.js +1 -1
  61. package/dist/entries/trends.cjs +2 -2
  62. package/dist/entries/trends.d.cts +11 -2
  63. package/dist/entries/trends.d.ts +11 -2
  64. package/dist/entries/trends.js +1 -1
  65. package/dist/entries/voices.cjs +2 -2
  66. package/dist/entries/voices.d.cts +16 -5
  67. package/dist/entries/voices.d.ts +16 -5
  68. package/dist/entries/voices.js +1 -1
  69. package/dist/index.cjs +22 -22
  70. package/dist/index.d.cts +2 -2
  71. package/dist/index.d.ts +2 -2
  72. package/dist/index.js +9 -9
  73. package/package.json +3 -3
  74. package/dist/chunk-3OARKXWT.cjs.map +0 -1
  75. package/dist/chunk-3QFTP5P5.js.map +0 -1
  76. package/dist/chunk-5LI4CVJB.cjs.map +0 -1
  77. package/dist/chunk-72ZJ72WF.js.map +0 -1
  78. package/dist/chunk-A42R7UT7.js.map +0 -1
  79. package/dist/chunk-GA4LZZKQ.cjs.map +0 -1
  80. package/dist/chunk-JK3YVG6A.cjs.map +0 -1
  81. package/dist/chunk-K2VWXAGK.js.map +0 -1
  82. package/dist/chunk-PTVUYLJO.cjs.map +0 -1
  83. package/dist/chunk-ZAATTX5W.js.map +0 -1
@@ -196,5 +196,5 @@ var VoicesResource = class {
196
196
  };
197
197
 
198
198
  exports.VoicesResource = VoicesResource;
199
- //# sourceMappingURL=chunk-R63TTR35.cjs.map
200
- //# sourceMappingURL=chunk-R63TTR35.cjs.map
199
+ //# sourceMappingURL=chunk-3CB4WYEY.cjs.map
200
+ //# sourceMappingURL=chunk-3CB4WYEY.cjs.map
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/resources/voices.ts"],"names":[],"mappings":";;;AAoBO,IAAM,iBAAN,MAAqB;AAAA,EAC1B,YAAoB,MAAA,EAA0B;AAA1B,IAAA,IAAA,CAAA,MAAA,GAAA,MAAA;AAAA,EAA2B;AAAA,EAA3B,MAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMpB,MAAM,IAAA,GAAyB;AAC7B,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAA6B,OAAO,YAAY,CAAA;AAC9E,IAAA,OAAO,GAAA,CAAI,MAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,aAAA,CAAc,MAAA,GAA6B,EAAC,EAAkC;AAC5E,IAAA,MAAM,QAA+D,EAAC;AACtE,IAAA,KAAA,MAAW,CAAC,CAAA,EAAG,CAAC,KAAK,MAAA,CAAO,OAAA,CAAQ,MAAM,CAAA,EAAG;AAC3C,MAAA,IAAI,CAAA,KAAM,UAAa,CAAA,KAAM,IAAA,IAAQ,MAAM,EAAA,EAAI,KAAA,CAAM,CAAC,CAAA,GAAI,CAAA;AAAA,IAC5D;AACA,IAAA,OAAO,KAAK,MAAA,CAAO,OAAA,CAAQ,OAAO,oBAAA,EAAsB,EAAE,OAAO,CAAA;AAAA,EACnE;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,UAAA,GAAoC;AACxC,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAAuC,OAAO,kBAAkB,CAAA;AAC9F,IAAA,OAAO,GAAA,CAAI,WAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,YAAY,KAAA,EAAgE;AAC1E,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAoB,MAAA,EAAQ,6BAA6B,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC7F;AAAA;AAAA,EAGA,MAAM,YAAY,EAAA,EAA2B;AAC3C,IAAA,MAAM,IAAA,CAAK,OAAO,OAAA,CAAc,QAAA,EAAU,oBAAoB,kBAAA,CAAmB,EAAE,CAAC,CAAA,CAAE,CAAA;AAAA,EACxF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAaA,OAAO,KAAA,EAewB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,qBAAqB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC5F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoCA,OAAO,KAAA,EAAyD;AAC9D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,yBAAyB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAChG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBA,QAAQ,KAAA,EAAoD;AAC1D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,iCAAiC,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACxG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAeA,UAAU,KAAA,EAAmD;AAC3D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,gCAAgC,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACvG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAO,KAAA,EAAqD;AAC1D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,oBAAoB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC3F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,KAAA,EAAoD;AACxD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,mBAAmB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC1F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,IAAI,KAAA,EAAiD;AACnD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,eAAe,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACtF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,eAAe,KAAA,EAcgB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,wBAAwB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC/F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,oBAAoB,KAAA,EAOI;AACtB,IAAA,MAAM,IAAA,GAAO,IAAI,QAAA,EAAS;AAG1B,IAAA,IAAA,CAAK,MAAA,CAAO,MAAA,EAAQ,KAAA,CAAM,IAAI,CAAA;AAC9B,IAAA,MAAM,OACJ,KAAA,CAAM,IAAA,YAAgB,IAAA,GAClB,KAAA,CAAM,OACN,IAAI,IAAA,CAAK,CAAC,KAAA,CAAM,IAAgB,CAAA,EAAG,EAAE,MAAM,KAAA,CAAM,WAAA,IAAe,cAAc,CAAA;AACpF,IAAA,IAAA,CAAK,MAAA,CAAO,MAAA,EAAQ,IAAA,EAAM,KAAA,CAAM,YAAY,QAAQ,CAAA;AACpD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAoB,MAAA,EAAQ,oBAAoB,EAAE,IAAA,EAAM,MAAM,CAAA;AAAA,EACnF;AACF","file":"chunk-R63TTR35.cjs","sourcesContent":["import type { Voice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from \"@nodaro/shared/pipeline\"\nimport type { AudioFxPreset } from \"@nodaro/shared/media\"\nimport type { DialogueProvider } from \"@nodaro/shared/models\"\nexport type { Voice, SharedVoice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from \"@nodaro/shared/pipeline\"\nexport type { AudioFxPreset } from \"@nodaro/shared/media\"\nimport type { NodaroClientCore } from \"../core.js\"\n\n/**\n * Re-export the shared voice types so SDK consumers don't have to add\n * `@nodaro/shared` as a second dependency just to type a `Voice` row, a\n * `VoiceClone`, or a `searchLibrary` call. Single source of truth lives in\n * `@nodaro/shared`.\n */\n/** Audio-FX preset union (reverb spaces / telephone / megaphone / echo / custom) — used by {@link VoiceChangerProInput.voiceFx}. */\n\n/**\n * Read access to ElevenLabs voices: the premade catalog plus the shared\n * community Voice Library (both public GETs, no body), and the signed-in\n * user's own pre-retirement voice clones (list / delete).\n */\nexport class VoicesResource {\n constructor(private client: NodaroClientCore) {}\n\n /**\n * List the premade ElevenLabs voices (`GET /v1/voices`). Falls back to a\n * curated set server-side when no ElevenLabs API key is configured.\n */\n async list(): Promise<Voice[]> {\n const res = await this.client.request<{ voices: Voice[] }>(\"GET\", \"/v1/voices\")\n return res.voices\n }\n\n /**\n * Search the shared/community Voice Library (`GET /v1/voices/library`). All\n * params are optional and forwarded as a querystring; `undefined` / `null` /\n * empty-string values are omitted so the server defaults apply. `hasMore`\n * drives \"load more\" pagination.\n */\n searchLibrary(params: VoiceLibraryParams = {}): Promise<VoiceLibraryResponse> {\n const query: Record<string, string | number | boolean | undefined> = {}\n for (const [k, v] of Object.entries(params)) {\n if (v !== undefined && v !== null && v !== \"\") query[k] = v as string | number | boolean\n }\n return this.client.request(\"GET\", \"/v1/voices/library\", { query })\n }\n\n /**\n * List the signed-in user's voice clones (`GET /v1/voice-clones`). The route\n * wraps the rows in `{ voiceClones }`; we unwrap to the bare array.\n */\n async listClones(): Promise<VoiceClone[]> {\n const res = await this.client.request<{ voiceClones: VoiceClone[] }>(\"GET\", \"/v1/voice-clones\")\n return res.voiceClones\n }\n\n /**\n * @deprecated Voice cloning is no longer offered on Nodaro (retired\n * 2026-09-15). The route answers `410 voice_cloning_retired` and this call\n * rejects with that error. Clones created before the retirement are still\n * listed by {@link VoicesResource.listClones} and still work as voice ids.\n * For a new custom voice use {@link VoicesResource.design}.\n */\n createClone(input: { name: string; audioUrl: string }): Promise<VoiceClone> {\n return this.client.request<VoiceClone>(\"POST\", \"/v1/voice-clones/from-url\", { body: input })\n }\n\n /** Delete one of the user's voice clones (`DELETE /v1/voice-clones/:id`). */\n async deleteClone(id: string): Promise<void> {\n await this.client.request<void>(\"DELETE\", `/v1/voice-clones/${encodeURIComponent(id)}`)\n }\n\n /**\n * Replace the voice in a recording — or in a whole talking video — with a\n * different voice (`POST /v1/voice-changer`). Pass `audioUrl` to revoice\n * audio→audio, or `videoUrl` to revoice an entire clip (the server demuxes\n * the audio, runs speech-to-speech, and remuxes onto the original video,\n * returning the video plus the new audio track). Exactly one of `audioUrl` /\n * `videoUrl` is required; when both are sent, video wins. `removeBackgroundNoise`\n * off keeps the music/SFX bed under the new voice; on yields a clean voice-only\n * result. Costs credits and runs async — poll `jobs.get(jobId)` for the result\n * (`output_data.videoUrl` + `output_data.audioUrl` in video mode).\n */\n change(input: {\n voiceId: string\n audioUrl?: string\n videoUrl?: string\n /** Speech-to-speech model id. Defaults to the server-configured default when omitted. */\n model?: string\n stability?: number\n similarityBoost?: number\n /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency/stability. */\n style?: number\n /** ElevenLabs speaker boost — sharpens fidelity to the target speaker (small latency cost). */\n useSpeakerBoost?: boolean\n /** Deterministic speech-to-speech seed (integer 0–4294967295) for reproducible output. Omit for random. */\n seed?: number\n removeBackgroundNoise?: boolean\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer\", { body: input })\n }\n\n /**\n * Recast each detected speaker in a multi-speaker recording to a different\n * voice (`POST /v1/voice-changer-pro`). `orderedVoices` maps speaker positions to\n * voices in detection order — speaker 0 → `orderedVoices[0]`, speaker 1 →\n * `orderedVoices[1]`, etc. Speakers beyond the end of `orderedVoices` keep\n * their original voice. Each entry is EITHER a bare voice id (premade name or\n * ElevenLabs UUID) OR a {@link VoiceChangerProVoice} object carrying per-voice\n * ElevenLabs speech-to-speech settings (stability / similarityBoost / style /\n * useSpeakerBoost / `seed`) plus a loudness `volumeMode` (and a manual\n * `volume`) — OR `null`, meaning keep this speaker's original voice (requires\n * a platform running cloud-plugins with keep-slot support). At least one\n * entry must be non-null. A per-voice `seed` makes that speaker's recast\n * reproducible.\n *\n * Pass `audioUrl` for audio-only recast or `videoUrl` to recast the audio\n * track of a video clip (the server demuxes, recasts, and remuxes).\n *\n * Voice and music are ALWAYS separated first — before recasting, the source\n * is split into an isolated vocal stem and a music/SFX stem.\n * `preserveBackground` (default `true`) only controls whether that\n * music/instrumental stem is mixed back under the new voices; set it `false`\n * for a clean voice-only result. `separationQuality` selects the quality of\n * the voice/music separation: `\"fast\"` (default, quicker — preserves more of\n * the voice) or `\"best\"` (finer voice/music separation).\n * `removeBackgroundNoise` additionally denoises the result.\n * `musicVolumeMode` sets the level of that preserved background (only relevant\n * when `preserveBackground` is on): `\"match\"` (default) keeps the original\n * level, `\"normalize\"` loudnorms it, `\"manual\"` uses `musicVolume`%.\n * `voiceFx` applies a reverb/echo to the COMBINED recast voices BEFORE the\n * background is mixed back in (effect sits on the voices, not the music bed).\n *\n * Cloud-only — costs credits and runs async; poll `jobs.get(jobId)` for the\n * result (`output_data.videoUrl` + `output_data.audioUrl` in video mode).\n */\n recast(input: VoiceChangerProInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro\", { body: input })\n }\n\n /**\n * Detect the speakers in a clip WITHOUT recasting yet\n * (`POST /v1/voice-changer-pro/analyze`) — the first step of the interactive\n * flow. Separates voice from music once and diarizes the vocals, returning the\n * speaker list so a user (or agent) can choose a voice per speaker before\n * committing to a paid recast. Poll `jobs.get(jobId)`: the completed job's\n * `output_data` carries the separated stem urls + the detected `speakers`\n * (each with `id`, time `segments`, `firstStartSec`, `wordCount`, `snippet`)\n * and the detected language — reshape it into a {@link VcpAnalysis} and pass it\n * as `recast({ ..., analysis })` to skip re-detection. With `suggestTitle`,\n * `output_data.suggestedTitle` also carries an LLM-proposed title.\n *\n * Cloud-only; costs credits and runs async.\n */\n analyze(input: VcpAnalyzeInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro/analyze\", { body: input })\n }\n\n /**\n * Render a final video from a mixed set of stems\n * (`POST /v1/voice-changer-pro/export`) — the last step of the interactive\n * flow. After `recast({ output: \"stems\" })` hands back the dry per-track stems\n * and the user has set levels / mutes / an effect in your editor, pass those\n * `tracks` (plus the source `videoUrl`) here to mix and remux into the finished\n * video. The video is stream-copied (never re-encoded), so the export is\n * bit-identical to your preview. At least one track must be un-muted (all-muted\n * is a 400); `voiceFx` is applied to the voice tracks at render time.\n *\n * Cloud-only; costs credits and runs async — poll `jobs.get(jobId)` for the\n * result (`output_data.videoUrl`).\n */\n exportMix(input: VcpExportInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro/export\", { body: input })\n }\n\n /**\n * Design a brand-new synthetic voice from a text description\n * (`POST /v1/voice-design`) — ElevenLabs text-to-voice. `text` (100–1000 chars)\n * is a preview line spoken in the designed voice; `voiceDescription` describes\n * the voice to create. Costs credits and runs async — poll `jobs.get(jobId)`\n * for the preview + the reusable voice id.\n */\n design(input: VoiceDesignInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-design\", { body: input })\n }\n\n /**\n * Generate speech in a voice described in natural language, without cloning\n * (`POST /v1/voice-remix`). `text` (1–5000 chars) is spoken in a voice matching\n * `voiceDescription`. Costs credits and runs async — poll `jobs.get(jobId)`.\n */\n remix(input: VoiceRemixInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-remix\", { body: input })\n }\n\n /**\n * Dub an audio clip into another language while preserving each speaker's voice\n * (`POST /v1/dubbing`). `targetLanguage` is an ISO code (e.g. `\"es\"`, `\"fr\"`);\n * `sourceLanguage` is auto-detected when omitted. Costs credits and runs async\n * — poll `jobs.get(jobId)`.\n */\n dub(input: DubbingInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/dubbing\", { body: input })\n }\n\n /**\n * Voice a multi-speaker script as ONE audio file\n * (`POST /v1/text-to-dialogue`) on ElevenLabs Dialogue v3 — the default — or\n * Dialogue v4 (`provider: \"elevenlabs-dialogue-v4\"`). Each `dialogue` line is\n * `{ text, voice }` in speaking order — `voice` is a premade voice name or an\n * ElevenLabs voice UUID (cloned/library voices work too; mixed casts are\n * fine). At most 5,000 characters total across lines on Dialogue v3 and\n * 10,000 on Dialogue v4 (≤2,000 recommended for best quality) and 10 unique\n * voices. Line text may\n * carry `[audio tags]` like `[laughs]`. Costs credits and runs async — poll\n * `jobs.get(jobId)` for `output_data.audioUrl` and — on every dialogue model,\n * since both return timings at no extra credits — `output_data.transcript`, a\n * `Transcript` (`@nodaro/shared`): `words[]` with `startMs` / `endMs` /\n * `speaker` and one `segments[]` entry per line, `speaker` being the line's\n * `voice` as you sent it. Pass `transcript.words` to\n * `addCaptions({ captions, autoTranscribe: false })` to caption the dialogue\n * without a second transcription.\n */\n textToDialogue(input: {\n dialogue: Array<{ text: string; voice: string }>\n /** Dialogue model: `\"elevenlabs-dialogue\"` (v3, the default) or `\"elevenlabs-dialogue-v4\"`. */\n provider?: DialogueProvider\n /** v3: exactly 0, 0.5 or 1 (0 = most variable, 1 = most stable). v4: any value 0–1. */\n stability?: number\n /** v4 only (v3 ignores it): how closely each line keeps its voice's character, 0–1. */\n similarityBoost?: number\n /** ISO 639-1 language hint (e.g. \"en\"); auto-detected when omitted. */\n languageCode?: string\n /** Deterministic sampling seed (integer 0–4294967295). Omit for random. */\n seed?: number\n /** Spell out numbers/dates/abbreviations: \"auto\" (default) | \"on\" | \"off\". */\n applyTextNormalization?: \"auto\" | \"on\" | \"off\"\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/text-to-dialogue\", { body: input })\n }\n\n /**\n * @deprecated Voice cloning is no longer offered on Nodaro (retired\n * 2026-09-15). The route answers `410 voice_cloning_retired` and this call\n * rejects with that error. See {@link VoicesResource.createClone}.\n */\n createCloneFromFile(input: {\n name: string\n file: Blob | Uint8Array | ArrayBuffer\n /** File name for the upload part (default `sample`). */\n filename?: string\n /** MIME type when `file` is a raw buffer (default `audio/mpeg`). */\n contentType?: string\n }): Promise<VoiceClone> {\n const form = new FormData()\n // Append the name field BEFORE the file so the route sees it in `fields`\n // (fastify-multipart's `req.file()` exposes fields parsed up to the file).\n form.append(\"name\", input.name)\n const blob =\n input.file instanceof Blob\n ? input.file\n : new Blob([input.file as BlobPart], { type: input.contentType ?? \"audio/mpeg\" })\n form.append(\"file\", blob, input.filename ?? \"sample\")\n return this.client.request<VoiceClone>(\"POST\", \"/v1/voice-clones\", { body: form })\n }\n}\n\n/**\n * One entry in {@link VoiceChangerProInput.orderedVoices}. Either a bare voice id\n * (premade name like `\"Rachel\"` or an ElevenLabs UUID for a custom clone), or\n * an object pinning per-voice speech-to-speech settings and the recast's\n * loudness behaviour for that speaker.\n */\nexport type VoiceChangerProVoice =\n | string\n | {\n /** Target voice — premade name (`\"Rachel\"`, `\"Aria\"`, …) or an ElevenLabs UUID for a custom clone. */\n voiceId: string\n /**\n * Which lane converts this speaker. `\"sts\"` (default) is the classic\n * speech-to-speech recast; `\"v3\"` and `\"v4\"` are RE-SPEAK — the\n * performance is REGENERATED from the transcript (supports `[audio\n * tags]`; original delivery is replaced, and lips won't match on video).\n * A Re-speak speaker needs transcript text: pass an `analysis` whose\n * `segments[].text` carries it (analyze now emits this), or omit\n * `analysis` and the engine re-speaks from its own transcription. For\n * `\"v3\"` (eleven_v3), `stability` accepts exactly 0 / 0.5 / 1, and\n * `similarityBoost` / `style` / `useSpeakerBoost` are ignored. For\n * `\"v4\"` (eleven_v4), `stability` takes any value 0–1 and\n * `similarityBoost` is honoured (`style` / `useSpeakerBoost` ignored);\n * each line is generated with its neighbouring lines as context for\n * smoother joins. Both Re-speak engines are priced the same, per started\n * 1,000 characters.\n */\n engine?: \"sts\" | \"v3\" | \"v4\"\n /** ElevenLabs stability (0–1). Higher = steadier, lower = more expressive. */\n stability?: number\n /** ElevenLabs similarity boost (0–1) — how closely the output hugs the target voice's timbre. */\n similarityBoost?: number\n /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency / stability. */\n style?: number\n /** ElevenLabs speaker boost — sharpens fidelity to the target speaker. */\n useSpeakerBoost?: boolean\n /**\n * Deterministic speech-to-speech seed (integer 0–4294967295) for\n * reproducible per-voice output — the same source + settings + seed\n * recast this speaker identically across runs. Omit for a random seed.\n */\n seed?: number\n /**\n * Loudness handling for this recast voice. `\"match\"` (default) matches the\n * original speaker's loudness; `\"normalize\"` applies loudness\n * normalization; `\"manual\"` uses `volume` as a percentage.\n */\n volumeMode?: \"match\" | \"normalize\" | \"manual\"\n /** Manual output volume as a percentage (0–200). Consulted only when `volumeMode === \"manual\"`. */\n volume?: number\n }\n\n/** Input for {@link VoicesResource.recast}. */\nexport interface VoiceChangerProInput {\n /** URL of an audio file to recast (audio → audio). Exactly one of `audioUrl` / `videoUrl` is required. */\n audioUrl?: string\n /** URL of a video file to recast (the audio track is recast and remuxed). Exactly one of `audioUrl` / `videoUrl` is required. */\n videoUrl?: string\n /**\n * Voices in speaker-detection order. Speaker N is mapped to `orderedVoices[N]`;\n * speakers beyond the array keep their original voice. Each entry is a bare\n * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings —\n * OR `null`, meaning keep this speaker's original voice (requires a platform\n * running cloud-plugins with keep-slot support). At least one entry must be\n * non-null.\n */\n orderedVoices: Array<VoiceChangerProVoice | null>\n /** Model to use for speech-to-speech. Defaults to the server-configured default when omitted. */\n model?: string\n /**\n * Mix the separated music / SFX stem back under the recast voices. Default\n * `true`. The voice is ALWAYS split out before recasting regardless of this\n * flag — `false` simply drops the music for a clean voice-only result.\n */\n preserveBackground?: boolean\n /**\n * Quality of the voice/music separation. `\"fast\"` (default, quicker —\n * preserves more of the voice) or `\"best\"` (finer voice/music separation).\n */\n separationQuality?: \"fast\" | \"best\"\n /** Strip background noise for a clean voice-only result. */\n removeBackgroundNoise?: boolean\n /**\n * Level of the preserved background music / SFX stem in the final mix. Only\n * relevant when `preserveBackground` is on (otherwise there is no background to\n * level). `\"match\"` (default) leaves the separated instrumental at its original\n * level; `\"normalize\"` applies loudness normalization; `\"manual\"` sets its\n * level to `musicVolume`%.\n */\n musicVolumeMode?: \"match\" | \"normalize\" | \"manual\"\n /** Background music level as a percentage (0–200). Consulted only when `musicVolumeMode === \"manual\"`. */\n musicVolume?: number\n /**\n * Node-level reverb/echo applied to the COMBINED recast voices **before** the\n * background is mixed back in (so the effect sits on the voices only, not the\n * music/SFX bed). Reverb presets (`room`, `hall`, `church`, …) use\n * `wetDryMix`; the `echo` / `custom` presets use `delayMs` + `decay`. Omit for\n * no effect.\n */\n voiceFx?: {\n /** Effect preset — reverb space, `telephone`, `megaphone`, `echo`, or `custom`. */\n preset: AudioFxPreset\n /** Reverb wet/dry mix as a percentage (0–100). Higher = wetter (more reverb). */\n wetDryMix?: number\n /** Echo delay in milliseconds (20–2000). Used by the `echo` / `custom` presets. */\n delayMs?: number\n /** Echo decay / feedback (0–1). Higher = more repeats. Used by the `echo` / `custom` presets. */\n decay?: number\n }\n /**\n * Output mode. `\"video\"` (default) mixes the recast voices with the preserved\n * background and returns a finished merged video. `\"stems\"` returns the dry,\n * unleveled per-track stems instead (rendering nothing) so you can drive an\n * INTERACTIVE mix — adjust levels/mutes/effect in your own UI, then render the\n * final video with {@link VoicesResource.exportMix}. This is how an app builds\n * a full editor around VCP rather than a one-shot recast.\n */\n output?: \"video\" | \"stems\"\n /**\n * A prior {@link VoicesResource.analyze} result. Pass it to SKIP re-detection:\n * the recast reuses the already-separated stems and speaker segments instead of\n * running separation + diarization again. This is the fast-path for the\n * detect → pick voices → recast interactive flow (analyze once, recast N times\n * as the user tweaks voice assignments). Omit to auto-detect from the source.\n */\n analysis?: VcpAnalysis\n}\n\n/** One detected speaker in a {@link VcpAnalysis} (from `analyze`). */\nexport interface VcpAnalysisSpeaker {\n /** Stable speaker id (first-appearance order). */\n id: string\n /** The speaker's spoken time ranges (seconds). `text` is what was said in\n * the range — the paid input for a Re-speak speaker (`engine: \"v3\"` or\n * `\"v4\"`; editable before conversion); the STS lane ignores it. */\n segments: Array<{ start: number; end: number; text?: string }>\n /** When the speaker first speaks (seconds). */\n firstStartSec?: number\n /** Rough word count across the clip — a proxy for how much this speaker says. */\n wordCount?: number\n /** The first few transcribed words, to help a user tell speakers apart. */\n snippet?: string\n}\n\n/**\n * The result of {@link VoicesResource.analyze}, reshaped to pass back into\n * {@link VoiceChangerProInput.analysis}. Read a completed analyze job's\n * `output_data` into this shape (it carries the separated stem urls + the\n * detected speakers) and thread it into `recast` to skip re-detection.\n */\nexport interface VcpAnalysis {\n /** URL of the isolated vocal stem. */\n vocalsUrl: string\n /** URL of the separated music/SFX stem (absent when the source had none). */\n backgroundUrl?: string\n /** The detected speakers, in first-appearance order. */\n speakers: VcpAnalysisSpeaker[]\n /** Scribe's detected language code, round-tripped so the recast auto-selects the STS model. */\n languageCode?: string\n /** Confidence (0–1) of {@link VcpAnalysis.languageCode}. */\n languageProbability?: number\n}\n\n/** Input for {@link VoicesResource.analyze}. */\nexport interface VcpAnalyzeInput {\n /** URL of an audio file to analyze. Exactly one of `audioUrl` / `videoUrl` is required. */\n audioUrl?: string\n /** URL of a video file to analyze (its audio track is used). Exactly one of `audioUrl` / `videoUrl` is required. */\n videoUrl?: string\n /** Quality of the voice/music separation run before diarization: `\"fast\"` (default) or `\"best\"`. */\n separationQuality?: \"fast\" | \"best\"\n /** Also suggest a conversion title from the transcript (returned on the job's `output_data.suggestedTitle`). */\n suggestTitle?: boolean\n}\n\n/** One track in a {@link VcpExportInput} mix. */\nexport interface VcpExportTrack {\n /** URL of the stem for this lane (a recast voice stem or the background stem). */\n url: string\n /** Fader position as a percentage: 0 = silent, 100 = unity, 200 = +6dB. */\n gain: number\n /** Whether this lane is muted in the mix. */\n muted: boolean\n /**\n * Which bucket the track is in, and so whether `voiceFx` lands on it. Defaults\n * to `\"voice\"`. Set `\"background\"` for the music/SFX lane (the effect never\n * touches it).\n */\n kind?: \"voice\" | \"background\"\n}\n\n/** Input for {@link VoicesResource.exportMix}. */\nexport interface VcpExportInput {\n /** The source video to remux the mixed audio onto (stream-copied — never re-encoded). */\n videoUrl: string\n /** The mix: one entry per lane. At least one must be un-muted (all-muted is a 400). Max 16 tracks. */\n tracks: VcpExportTrack[]\n /**\n * A reverb/echo applied to the VOICE tracks only (not `\"background\"` lanes)\n * at render time — so iterating the effect in your editor is free until you\n * export. Same shape as {@link VoiceChangerProInput.voiceFx}.\n */\n voiceFx?: VoiceChangerProInput[\"voiceFx\"]\n}\n\n/** Input for {@link VoicesResource.design}. */\nexport interface VoiceDesignInput {\n /** A preview line (100–1000 chars) spoken in the designed voice. */\n text: string\n /** Natural-language description of the voice to create. */\n voiceDescription: string\n /** Voice-design model id. Defaults to the server-configured default. */\n model?: string\n /** Output loudness (-1..1). */\n loudness?: number\n /** How strongly the description steers the design (0–100). */\n guidanceScale?: number\n /** Deterministic seed for a reproducible design. */\n seed?: number\n /** Design quality knob (provider-specific). */\n quality?: number\n /** Enhance the generated voice. */\n shouldEnhance?: boolean\n /** Optional extra prompt context (≤8000 chars). */\n userPrompt?: string\n}\n\n/** Input for {@link VoicesResource.remix}. */\nexport interface VoiceRemixInput {\n /** The text (1–5000 chars) to speak in the described voice. */\n text: string\n /** Natural-language description of the voice to speak in. */\n voiceDescription: string\n /** Optional extra prompt context (≤8000 chars). */\n userPrompt?: string\n}\n\n/** Input for {@link VoicesResource.dub}. Provide exactly ONE source:\n * `audioUrl` (dubbed audio out), `videoUrl` (the dubbed VIDEO out, plus the\n * dubbed audio track as `output_data.audioUrl`), or `sourceUrl` (a public\n * YouTube/TikTok/direct link ElevenLabs fetches itself). Priced per minute\n * of the dubbed span; the span is capped at 30 minutes — use\n * `startTime`/`endTime` to dub part of a longer source. */\nexport interface DubbingInput {\n /** URL of the audio to dub (audio mode). */\n audioUrl?: string\n /** URL of the video to dub — the result is the dubbed VIDEO (+ audio track). */\n videoUrl?: string\n /** Public page/media URL (YouTube, TikTok, or a direct link) — ElevenLabs fetches it directly. */\n sourceUrl?: string\n /** Target language ISO code (2–10 chars), e.g. `\"es\"`, `\"pt-BR\"`. */\n targetLanguage: string\n /** Source language ISO code; auto-detected when omitted. */\n sourceLanguage?: string\n /** Expected number of speakers (1–20; 0 = auto-detect) — improves separation when known. */\n numSpeakers?: number\n /** Keep the original voices instead of cloning them into the target language. */\n disableVoiceCloning?: boolean\n /** Drop the background/music bed from the dubbed output. */\n dropBackgroundAudio?: boolean\n /** Dub only this window of the source (seconds). */\n startTime?: number\n endTime?: number\n /** Keep the source resolution on video dubs (slower render). */\n highestResolution?: boolean\n useProfanityFilter?: boolean\n /** Experimental: steer dubbed voices toward an accent. */\n targetAccent?: string\n /** Apply ElevenLabs' own watermark to video dubs. */\n watermark?: boolean\n}\n"]}
1
+ {"version":3,"sources":["../src/resources/voices.ts"],"names":[],"mappings":";;;AAoBO,IAAM,iBAAN,MAAqB;AAAA,EAC1B,YAAoB,MAAA,EAA0B;AAA1B,IAAA,IAAA,CAAA,MAAA,GAAA,MAAA;AAAA,EAA2B;AAAA,EAA3B,MAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMpB,MAAM,IAAA,GAAyB;AAC7B,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAA6B,OAAO,YAAY,CAAA;AAC9E,IAAA,OAAO,GAAA,CAAI,MAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,aAAA,CAAc,MAAA,GAA6B,EAAC,EAAkC;AAC5E,IAAA,MAAM,QAA+D,EAAC;AACtE,IAAA,KAAA,MAAW,CAAC,CAAA,EAAG,CAAC,KAAK,MAAA,CAAO,OAAA,CAAQ,MAAM,CAAA,EAAG;AAC3C,MAAA,IAAI,CAAA,KAAM,UAAa,CAAA,KAAM,IAAA,IAAQ,MAAM,EAAA,EAAI,KAAA,CAAM,CAAC,CAAA,GAAI,CAAA;AAAA,IAC5D;AACA,IAAA,OAAO,KAAK,MAAA,CAAO,OAAA,CAAQ,OAAO,oBAAA,EAAsB,EAAE,OAAO,CAAA;AAAA,EACnE;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,UAAA,GAAoC;AACxC,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAAuC,OAAO,kBAAkB,CAAA;AAC9F,IAAA,OAAO,GAAA,CAAI,WAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,YAAY,KAAA,EAAgE;AAC1E,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAoB,MAAA,EAAQ,6BAA6B,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC7F;AAAA;AAAA,EAGA,MAAM,YAAY,EAAA,EAA2B;AAC3C,IAAA,MAAM,IAAA,CAAK,OAAO,OAAA,CAAc,QAAA,EAAU,oBAAoB,kBAAA,CAAmB,EAAE,CAAC,CAAA,CAAE,CAAA;AAAA,EACxF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAaA,OAAO,KAAA,EAewB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,qBAAqB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC5F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoCA,OAAO,KAAA,EAAyD;AAC9D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,yBAAyB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAChG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBA,QAAQ,KAAA,EAAoD;AAC1D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,iCAAiC,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACxG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAeA,UAAU,KAAA,EAAmD;AAC3D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,gCAAgC,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACvG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAO,KAAA,EAAqD;AAC1D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,oBAAoB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC3F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,KAAA,EAAoD;AACxD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,mBAAmB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC1F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,IAAI,KAAA,EAAiD;AACnD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,eAAe,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACtF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,eAAe,KAAA,EAcgB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,wBAAwB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC/F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,oBAAoB,KAAA,EAOI;AACtB,IAAA,MAAM,IAAA,GAAO,IAAI,QAAA,EAAS;AAG1B,IAAA,IAAA,CAAK,MAAA,CAAO,MAAA,EAAQ,KAAA,CAAM,IAAI,CAAA;AAC9B,IAAA,MAAM,OACJ,KAAA,CAAM,IAAA,YAAgB,IAAA,GAClB,KAAA,CAAM,OACN,IAAI,IAAA,CAAK,CAAC,KAAA,CAAM,IAAgB,CAAA,EAAG,EAAE,MAAM,KAAA,CAAM,WAAA,IAAe,cAAc,CAAA;AACpF,IAAA,IAAA,CAAK,MAAA,CAAO,MAAA,EAAQ,IAAA,EAAM,KAAA,CAAM,YAAY,QAAQ,CAAA;AACpD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAoB,MAAA,EAAQ,oBAAoB,EAAE,IAAA,EAAM,MAAM,CAAA;AAAA,EACnF;AACF","file":"chunk-3CB4WYEY.cjs","sourcesContent":["import type { Voice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from \"@nodaro/shared/pipeline\"\nimport type { AudioFxPreset } from \"@nodaro/shared/media\"\nimport type { DialogueProvider } from \"@nodaro/shared/models\"\nexport type { Voice, SharedVoice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from \"@nodaro/shared/pipeline\"\nexport type { AudioFxPreset } from \"@nodaro/shared/media\"\nimport type { NodaroClientCore } from \"../core.js\"\n\n/**\n * Re-export the shared voice types so SDK consumers don't have to add\n * `@nodaro/shared` as a second dependency just to type a `Voice` row, a\n * `VoiceClone`, or a `searchLibrary` call. Single source of truth lives in\n * `@nodaro/shared`.\n */\n/** Audio-FX preset union (reverb spaces / telephone / megaphone / echo / custom) — used by {@link VoiceChangerProInput.voiceFx}. */\n\n/**\n * Read access to ElevenLabs voices: the premade catalog plus the shared\n * community Voice Library (both public GETs, no body), and the signed-in\n * user's own pre-retirement voice clones (list / delete).\n */\nexport class VoicesResource {\n constructor(private client: NodaroClientCore) {}\n\n /**\n * List the premade ElevenLabs voices (`GET /v1/voices`). Falls back to a\n * curated set server-side when no ElevenLabs API key is configured.\n */\n async list(): Promise<Voice[]> {\n const res = await this.client.request<{ voices: Voice[] }>(\"GET\", \"/v1/voices\")\n return res.voices\n }\n\n /**\n * Search the shared/community Voice Library (`GET /v1/voices/library`). All\n * params are optional and forwarded as a querystring; `undefined` / `null` /\n * empty-string values are omitted so the server defaults apply. `hasMore`\n * drives \"load more\" pagination.\n */\n searchLibrary(params: VoiceLibraryParams = {}): Promise<VoiceLibraryResponse> {\n const query: Record<string, string | number | boolean | undefined> = {}\n for (const [k, v] of Object.entries(params)) {\n if (v !== undefined && v !== null && v !== \"\") query[k] = v as string | number | boolean\n }\n return this.client.request(\"GET\", \"/v1/voices/library\", { query })\n }\n\n /**\n * List the signed-in user's voice clones (`GET /v1/voice-clones`). The route\n * wraps the rows in `{ voiceClones }`; we unwrap to the bare array.\n */\n async listClones(): Promise<VoiceClone[]> {\n const res = await this.client.request<{ voiceClones: VoiceClone[] }>(\"GET\", \"/v1/voice-clones\")\n return res.voiceClones\n }\n\n /**\n * @deprecated Voice cloning is no longer offered on Nodaro (retired\n * 2026-09-15). The route answers `410 voice_cloning_retired` and this call\n * rejects with that error. Clones created before the retirement are still\n * listed by {@link VoicesResource.listClones} and still work as voice ids.\n * For a new custom voice use {@link VoicesResource.design}.\n */\n createClone(input: { name: string; audioUrl: string }): Promise<VoiceClone> {\n return this.client.request<VoiceClone>(\"POST\", \"/v1/voice-clones/from-url\", { body: input })\n }\n\n /** Delete one of the user's voice clones (`DELETE /v1/voice-clones/:id`). */\n async deleteClone(id: string): Promise<void> {\n await this.client.request<void>(\"DELETE\", `/v1/voice-clones/${encodeURIComponent(id)}`)\n }\n\n /**\n * Replace the voice in a recording — or in a whole talking video — with a\n * different voice (`POST /v1/voice-changer`). Pass `audioUrl` to revoice\n * audio→audio, or `videoUrl` to revoice an entire clip (the server demuxes\n * the audio, runs speech-to-speech, and remuxes onto the original video,\n * returning the video plus the new audio track). Exactly one of `audioUrl` /\n * `videoUrl` is required; when both are sent, video wins. `removeBackgroundNoise`\n * off keeps the music/SFX bed under the new voice; on yields a clean voice-only\n * result. Costs credits and runs async — poll `jobs.get(jobId)` for the result\n * (`output_data.videoUrl` + `output_data.audioUrl` in video mode).\n */\n change(input: {\n voiceId: string\n audioUrl?: string\n videoUrl?: string\n /** Speech-to-speech model id. Defaults to the server-configured default when omitted. */\n model?: string\n stability?: number\n similarityBoost?: number\n /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency/stability. */\n style?: number\n /** ElevenLabs speaker boost — sharpens fidelity to the target speaker (small latency cost). */\n useSpeakerBoost?: boolean\n /** Deterministic speech-to-speech seed (integer 0–4294967295) for reproducible output. Omit for random. */\n seed?: number\n removeBackgroundNoise?: boolean\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer\", { body: input })\n }\n\n /**\n * Recast each detected speaker in a multi-speaker recording to a different\n * voice (`POST /v1/voice-changer-pro`). `orderedVoices` maps speaker positions to\n * voices in detection order — speaker 0 → `orderedVoices[0]`, speaker 1 →\n * `orderedVoices[1]`, etc. Speakers beyond the end of `orderedVoices` keep\n * their original voice. Each entry is EITHER a bare voice id (premade name or\n * ElevenLabs UUID) OR a {@link VoiceChangerProVoice} object carrying per-voice\n * ElevenLabs speech-to-speech settings (stability / similarityBoost / style /\n * useSpeakerBoost / `seed`) plus a loudness `volumeMode` (and a manual\n * `volume`) — OR `null`, meaning keep this speaker's original voice (requires\n * a platform running cloud-plugins with keep-slot support). At least one\n * entry must be non-null. A per-voice `seed` makes that speaker's recast\n * reproducible.\n *\n * Pass `audioUrl` for audio-only recast or `videoUrl` to recast the audio\n * track of a video clip (the server demuxes, recasts, and remuxes).\n *\n * Voice and music are ALWAYS separated first — before recasting, the source\n * is split into an isolated vocal stem and a music/SFX stem.\n * `preserveBackground` (default `true`) only controls whether that\n * music/instrumental stem is mixed back under the new voices; set it `false`\n * for a clean voice-only result. `separationQuality` selects the quality of\n * the voice/music separation: `\"fast\"` (default, quicker — preserves more of\n * the voice) or `\"best\"` (finer voice/music separation).\n * `removeBackgroundNoise` additionally denoises the result.\n * `musicVolumeMode` sets the level of that preserved background (only relevant\n * when `preserveBackground` is on): `\"match\"` (default) keeps the original\n * level, `\"normalize\"` loudnorms it, `\"manual\"` uses `musicVolume`%.\n * `voiceFx` applies a reverb/echo to the COMBINED recast voices BEFORE the\n * background is mixed back in (effect sits on the voices, not the music bed).\n *\n * Cloud-only — costs credits and runs async; poll `jobs.get(jobId)` for the\n * result (`output_data.videoUrl` + `output_data.audioUrl` in video mode).\n */\n recast(input: VoiceChangerProInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro\", { body: input })\n }\n\n /**\n * Detect the speakers in a clip WITHOUT recasting yet\n * (`POST /v1/voice-changer-pro/analyze`) — the first step of the interactive\n * flow. Separates voice from music once and diarizes the vocals, returning the\n * speaker list so a user (or agent) can choose a voice per speaker before\n * committing to a paid recast. Poll `jobs.get(jobId)`: the completed job's\n * `output_data` carries the separated stem urls + the detected `speakers`\n * (each with `id`, time `segments`, `firstStartSec`, `wordCount`, `snippet`)\n * and the detected language — reshape it into a {@link VcpAnalysis} and pass it\n * as `recast({ ..., analysis })` to skip re-detection. With `suggestTitle`,\n * `output_data.suggestedTitle` also carries an LLM-proposed title.\n *\n * Cloud-only; costs credits and runs async.\n */\n analyze(input: VcpAnalyzeInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro/analyze\", { body: input })\n }\n\n /**\n * Render a final video from a mixed set of stems\n * (`POST /v1/voice-changer-pro/export`) — the last step of the interactive\n * flow. After `recast({ output: \"stems\" })` hands back the dry per-track stems\n * and the user has set levels / mutes / an effect in your editor, pass those\n * `tracks` (plus the source `videoUrl`) here to mix and remux into the finished\n * video. The video is stream-copied (never re-encoded), so the export is\n * bit-identical to your preview. At least one track must be un-muted (all-muted\n * is a 400); `voiceFx` is applied to the voice tracks at render time.\n *\n * Cloud-only; costs credits and runs async — poll `jobs.get(jobId)` for the\n * result (`output_data.videoUrl`).\n */\n exportMix(input: VcpExportInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro/export\", { body: input })\n }\n\n /**\n * Design a brand-new synthetic voice from a text description\n * (`POST /v1/voice-design`) — ElevenLabs text-to-voice. `text` (100–1000 chars)\n * is a preview line spoken in the designed voice; `voiceDescription` describes\n * the voice to create. Costs credits and runs async — poll `jobs.get(jobId)`\n * for the preview + the reusable voice id.\n */\n design(input: VoiceDesignInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-design\", { body: input })\n }\n\n /**\n * Generate speech in a voice described in natural language, without cloning\n * (`POST /v1/voice-remix`). `text` (1–5000 chars) is spoken in a voice matching\n * `voiceDescription`. Costs credits and runs async — poll `jobs.get(jobId)`.\n */\n remix(input: VoiceRemixInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-remix\", { body: input })\n }\n\n /**\n * Dub an audio clip into another language while preserving each speaker's voice\n * (`POST /v1/dubbing`). `targetLanguage` is an ISO code (e.g. `\"es\"`, `\"fr\"`);\n * `sourceLanguage` is auto-detected when omitted. Costs credits and runs async\n * — poll `jobs.get(jobId)`.\n */\n dub(input: DubbingInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/dubbing\", { body: input })\n }\n\n /**\n * Voice a multi-speaker script as ONE audio file\n * (`POST /v1/text-to-dialogue`) on ElevenLabs Dialogue v3 — the default — or\n * Dialogue v4 (`provider: \"elevenlabs-dialogue-v4\"`). Each `dialogue` line is\n * `{ text, voice }` in speaking order — `voice` is a premade voice name or an\n * ElevenLabs voice UUID (cloned/library voices work too; mixed casts are\n * fine). At most 5,000 characters total across lines on Dialogue v3 and\n * 10,000 on Dialogue v4 (≤2,000 recommended for best quality) and 10 unique\n * voices. Line text may\n * carry `[audio tags]` like `[laughs]`. Costs credits and runs async — poll\n * `jobs.get(jobId)` for `output_data.audioUrl` and — on every dialogue model,\n * since both return timings at no extra credits — `output_data.transcript`, a\n * `Transcript` (`@nodaro/shared`): `words[]` with `startMs` / `endMs` /\n * `speaker` and one `segments[]` entry per line, `speaker` being the line's\n * `voice` as you sent it. Pass `transcript.words` to\n * `addCaptions({ captions, autoTranscribe: false })` to caption the dialogue\n * without a second transcription.\n */\n textToDialogue(input: {\n dialogue: Array<{ text: string; voice: string }>\n /** Dialogue model: `\"elevenlabs-dialogue\"` (v3, the default) or `\"elevenlabs-dialogue-v4\"`. */\n provider?: DialogueProvider\n /** v3: exactly 0, 0.5 or 1 (0 = most variable, 1 = most stable). v4: any value 0–1. */\n stability?: number\n /** v4 only (v3 ignores it): how closely each line keeps its voice's character, 0–1. */\n similarityBoost?: number\n /** ISO 639-1 language hint (e.g. \"en\"); auto-detected when omitted. */\n languageCode?: string\n /** Deterministic sampling seed (integer 0–4294967295). Omit for random. */\n seed?: number\n /** Spell out numbers/dates/abbreviations: \"auto\" (default) | \"on\" | \"off\". */\n applyTextNormalization?: \"auto\" | \"on\" | \"off\"\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/text-to-dialogue\", { body: input })\n }\n\n /**\n * @deprecated Voice cloning is no longer offered on Nodaro (retired\n * 2026-09-15). The route answers `410 voice_cloning_retired` and this call\n * rejects with that error. See {@link VoicesResource.createClone}.\n */\n createCloneFromFile(input: {\n name: string\n file: Blob | Uint8Array | ArrayBuffer\n /** File name for the upload part (default `sample`). */\n filename?: string\n /** MIME type when `file` is a raw buffer (default `audio/mpeg`). */\n contentType?: string\n }): Promise<VoiceClone> {\n const form = new FormData()\n // Append the name field BEFORE the file so the route sees it in `fields`\n // (fastify-multipart's `req.file()` exposes fields parsed up to the file).\n form.append(\"name\", input.name)\n const blob =\n input.file instanceof Blob\n ? input.file\n : new Blob([input.file as BlobPart], { type: input.contentType ?? \"audio/mpeg\" })\n form.append(\"file\", blob, input.filename ?? \"sample\")\n return this.client.request<VoiceClone>(\"POST\", \"/v1/voice-clones\", { body: form })\n }\n}\n\n/**\n * One entry in {@link VoiceChangerProInput.orderedVoices}. Either a bare voice id\n * (premade name like `\"Rachel\"` or an ElevenLabs UUID for a custom clone), or\n * an object pinning per-voice speech-to-speech settings and the recast's\n * loudness behaviour for that speaker.\n */\nexport type VoiceChangerProVoice =\n | string\n | {\n /** Target voice — premade name (`\"Rachel\"`, `\"Aria\"`, …) or an ElevenLabs UUID for a custom clone. */\n voiceId: string\n /**\n * Which lane converts this speaker. `\"sts\"` (default) is the classic\n * speech-to-speech recast; `\"v3\"` and `\"v4\"` are RE-SPEAK — the\n * performance is REGENERATED from the transcript (supports `[audio\n * tags]`; original delivery is replaced, and lips won't match on video).\n * A Re-speak speaker needs transcript text: pass an `analysis` whose\n * `segments[].text` carries it (analyze now emits this), or omit\n * `analysis` and the engine re-speaks from its own transcription. For\n * `\"v3\"` (eleven_v3), `stability` accepts exactly 0 / 0.5 / 1, and\n * `similarityBoost` / `style` / `useSpeakerBoost` are ignored. For\n * `\"v4\"` (eleven_v4), `stability` takes any value 0–1 and\n * `similarityBoost` is honoured (`style` / `useSpeakerBoost` ignored);\n * each line is generated with its neighbouring lines as context for\n * smoother joins. Both Re-speak engines are priced the same, per started\n * 1,000 characters.\n */\n engine?: \"sts\" | \"v3\" | \"v4\"\n /** ElevenLabs stability (0–1). Higher = steadier, lower = more expressive. */\n stability?: number\n /** ElevenLabs similarity boost (0–1) — how closely the output hugs the target voice's timbre. */\n similarityBoost?: number\n /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency / stability. */\n style?: number\n /** ElevenLabs speaker boost — sharpens fidelity to the target speaker. */\n useSpeakerBoost?: boolean\n /**\n * Deterministic speech-to-speech seed (integer 0–4294967295) for\n * reproducible per-voice output — the same source + settings + seed\n * recast this speaker identically across runs. Omit for a random seed.\n */\n seed?: number\n /**\n * Loudness handling for this recast voice. `\"match\"` (default) matches the\n * original speaker's loudness; `\"normalize\"` applies loudness\n * normalization; `\"manual\"` uses `volume` as a percentage.\n */\n volumeMode?: \"match\" | \"normalize\" | \"manual\"\n /** Manual output volume as a percentage (0–200). Consulted only when `volumeMode === \"manual\"`. */\n volume?: number\n }\n\n/** Input for {@link VoicesResource.recast}. */\nexport interface VoiceChangerProInput {\n /** URL of an audio file to recast (audio → audio). Exactly one of `audioUrl` / `videoUrl` is required. */\n audioUrl?: string\n /** URL of a video file to recast (the audio track is recast and remuxed). Exactly one of `audioUrl` / `videoUrl` is required. */\n videoUrl?: string\n /**\n * Voices in speaker-detection order. Speaker N is mapped to `orderedVoices[N]`;\n * speakers beyond the array keep their original voice. Each entry is a bare\n * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings —\n * OR `null`, meaning keep this speaker's original voice (requires a platform\n * running cloud-plugins with keep-slot support). At least one entry must be\n * non-null.\n */\n orderedVoices: Array<VoiceChangerProVoice | null>\n /** Model to use for speech-to-speech. Defaults to the server-configured default when omitted. */\n model?: string\n /**\n * Mix the separated music / SFX stem back under the recast voices. Default\n * `true`. The voice is ALWAYS split out before recasting regardless of this\n * flag — `false` simply drops the music for a clean voice-only result.\n */\n preserveBackground?: boolean\n /**\n * Quality of the voice/music separation. `\"fast\"` (default, quicker —\n * preserves more of the voice) or `\"best\"` (finer voice/music separation).\n */\n separationQuality?: \"fast\" | \"best\"\n /** Strip background noise for a clean voice-only result. */\n removeBackgroundNoise?: boolean\n /**\n * Level of the preserved background music / SFX stem in the final mix. Only\n * relevant when `preserveBackground` is on (otherwise there is no background to\n * level). `\"match\"` (default) leaves the separated instrumental at its original\n * level; `\"normalize\"` applies loudness normalization; `\"manual\"` sets its\n * level to `musicVolume`%.\n */\n musicVolumeMode?: \"match\" | \"normalize\" | \"manual\"\n /** Background music level as a percentage (0–200). Consulted only when `musicVolumeMode === \"manual\"`. */\n musicVolume?: number\n /**\n * Node-level reverb/echo applied to the COMBINED recast voices **before** the\n * background is mixed back in (so the effect sits on the voices only, not the\n * music/SFX bed). Reverb presets (`room`, `hall`, `church`, …) use\n * `wetDryMix`; the `echo` / `custom` presets use `delayMs` + `decay`. Omit for\n * no effect.\n */\n voiceFx?: {\n /** Effect preset — reverb space, `telephone`, `megaphone`, `echo`, or `custom`. */\n preset: AudioFxPreset\n /** Reverb wet/dry mix as a percentage (0–100). Higher = wetter (more reverb). */\n wetDryMix?: number\n /** Echo delay in milliseconds (20–2000). Used by the `echo` / `custom` presets. */\n delayMs?: number\n /** Echo decay / feedback (0–1). Higher = more repeats. Used by the `echo` / `custom` presets. */\n decay?: number\n }\n /**\n * Output mode. `\"video\"` (default) mixes the recast voices with the preserved\n * background and returns a finished merged video. `\"stems\"` returns the dry,\n * unleveled per-track stems instead (rendering nothing) so you can drive an\n * INTERACTIVE mix — adjust levels/mutes/effect in your own UI, then render the\n * final video with {@link VoicesResource.exportMix}. This is how an app builds\n * a full editor around VCP rather than a one-shot recast.\n */\n output?: \"video\" | \"stems\"\n /**\n * A prior {@link VoicesResource.analyze} result. Pass it to SKIP re-detection:\n * the recast reuses the already-separated stems and speaker segments instead of\n * running separation + diarization again. This is the fast-path for the\n * detect → pick voices → recast interactive flow (analyze once, recast N times\n * as the user tweaks voice assignments). Omit to auto-detect from the source.\n */\n analysis?: VcpAnalysis\n}\n\n/** One detected speaker in a {@link VcpAnalysis} (from `analyze`). */\nexport interface VcpAnalysisSpeaker {\n /** Stable speaker id (first-appearance order). */\n id: string\n /** The speaker's spoken time ranges (seconds). `text` is what was said in\n * the range — the paid input for a Re-speak speaker (`engine: \"v3\"` or\n * `\"v4\"`; editable before conversion); the STS lane ignores it. */\n segments: Array<{ start: number; end: number; text?: string }>\n /** When the speaker first speaks (seconds). */\n firstStartSec?: number\n /** Rough word count across the clip — a proxy for how much this speaker says. */\n wordCount?: number\n /** The first few transcribed words, to help a user tell speakers apart. */\n snippet?: string\n}\n\n/**\n * The result of {@link VoicesResource.analyze}, reshaped to pass back into\n * {@link VoiceChangerProInput.analysis}. Read a completed analyze job's\n * `output_data` into this shape (it carries the separated stem urls + the\n * detected speakers) and thread it into `recast` to skip re-detection.\n */\nexport interface VcpAnalysis {\n /** URL of the isolated vocal stem. */\n vocalsUrl: string\n /** URL of the separated music/SFX stem (absent when the source had none). */\n backgroundUrl?: string\n /** The detected speakers, in first-appearance order. */\n speakers: VcpAnalysisSpeaker[]\n /** Scribe's detected language code, round-tripped so the recast auto-selects the STS model. */\n languageCode?: string\n /** Confidence (0–1) of {@link VcpAnalysis.languageCode}. */\n languageProbability?: number\n}\n\n/** Input for {@link VoicesResource.analyze}. */\nexport interface VcpAnalyzeInput {\n /** URL of an audio file to analyze. Exactly one of `audioUrl` / `videoUrl` is required. */\n audioUrl?: string\n /** URL of a video file to analyze (its audio track is used). Exactly one of `audioUrl` / `videoUrl` is required. */\n videoUrl?: string\n /** Quality of the voice/music separation run before diarization: `\"fast\"` (default) or `\"best\"`. */\n separationQuality?: \"fast\" | \"best\"\n /** Also suggest a conversion title from the transcript (returned on the job's `output_data.suggestedTitle`). */\n suggestTitle?: boolean\n}\n\n/** One track in a {@link VcpExportInput} mix. */\nexport interface VcpExportTrack {\n /** URL of the stem for this lane (a recast voice stem or the background stem). */\n url: string\n /** Fader position as a percentage: 0 = silent, 100 = unity, 200 = +6dB. */\n gain: number\n /** Whether this lane is muted in the mix. */\n muted: boolean\n /**\n * Which bucket the track is in, and so whether `voiceFx` lands on it. Defaults\n * to `\"voice\"`. Set `\"background\"` for the music/SFX lane (the effect never\n * touches it).\n */\n kind?: \"voice\" | \"background\"\n}\n\n/** Input for {@link VoicesResource.exportMix}. */\nexport interface VcpExportInput {\n /** The source video to remux the mixed audio onto (stream-copied — never re-encoded). */\n videoUrl: string\n /** The mix: one entry per lane. At least one must be un-muted (all-muted is a 400). Max 16 tracks. */\n tracks: VcpExportTrack[]\n /**\n * A reverb/echo applied to the VOICE tracks only (not `\"background\"` lanes)\n * at render time — so iterating the effect in your editor is free until you\n * export. Same shape as {@link VoiceChangerProInput.voiceFx}.\n */\n voiceFx?: VoiceChangerProInput[\"voiceFx\"]\n}\n\n/** Input for {@link VoicesResource.design}. */\nexport interface VoiceDesignInput {\n /** A preview line (100–1000 chars) spoken in the designed voice. */\n text: string\n /** Natural-language description of the voice to create. */\n voiceDescription: string\n /** Voice-design model id. Defaults to the server-configured default. */\n model?: string\n /** Output loudness (-1..1). */\n loudness?: number\n /** How strongly the description steers the design (0–100). */\n guidanceScale?: number\n /** Deterministic seed for a reproducible design. */\n seed?: number\n /** Design quality knob (provider-specific). */\n quality?: number\n /** Enhance the generated voice. */\n shouldEnhance?: boolean\n /** Optional extra prompt context (≤8000 chars). */\n userPrompt?: string\n}\n\n/** Input for {@link VoicesResource.remix}. */\nexport interface VoiceRemixInput {\n /** The text (1–5000 chars) to speak in the described voice. */\n text: string\n /** Natural-language description of the voice to speak in. */\n voiceDescription: string\n /** Optional extra prompt context (≤8000 chars). */\n userPrompt?: string\n}\n\n/** Input for {@link VoicesResource.dub}. Provide exactly ONE source:\n * `audioUrl` (dubbed audio out), `videoUrl` (the dubbed VIDEO out, plus the\n * dubbed audio track as `output_data.audioUrl`), or `sourceUrl` (a link to\n * a video's page, or a direct link to a file, that the dubbing service\n * fetches itself). Priced per minute of the dubbed span; the span is capped\n * at 30 minutes — use `startTime`/`endTime` to dub part of a longer source. */\nexport interface DubbingInput {\n /** URL of the audio to dub (audio mode). */\n audioUrl?: string\n /** URL of the video to dub — the result is the dubbed VIDEO (+ audio track). */\n videoUrl?: string\n /**\n * A link to a video's page, or a direct link to a file, that the dubbing\n * service fetches itself: an import of that link, made under the content\n * rights declaration (see `confirmationId`).\n */\n sourceUrl?: string\n /**\n * The content rights confirmation for `sourceUrl` (`contentRights.confirm()`).\n * Without one, a deployment that requires a confirmation refuses the dub with\n * `403`, unless the request comes through a key or an agent whose account\n * holds a standing authorization.\n */\n confirmationId?: string\n /** Target language ISO code (2–10 chars), e.g. `\"es\"`, `\"pt-BR\"`. */\n targetLanguage: string\n /** Source language ISO code; auto-detected when omitted. */\n sourceLanguage?: string\n /** Expected number of speakers (1–20; 0 = auto-detect) — improves separation when known. */\n numSpeakers?: number\n /** Keep the original voices instead of cloning them into the target language. */\n disableVoiceCloning?: boolean\n /** Drop the background/music bed from the dubbed output. */\n dropBackgroundAudio?: boolean\n /** Dub only this window of the source (seconds). */\n startTime?: number\n endTime?: number\n /** Keep the source resolution on video dubs (slower render). */\n highestResolution?: boolean\n useProfanityFilter?: boolean\n /** Experimental: steer dubbed voices toward an accent. */\n targetAccent?: string\n /** Apply ElevenLabs' own watermark to video dubs. */\n watermark?: boolean\n}\n"]}
@@ -185,13 +185,28 @@ var MediaResource = class {
185
185
  * short leftover is evened out with the parts before it. Video is cut on
186
186
  * whole frames with each part's sound matching its picture; audio parts are
187
187
  * sample-exact, in `audioFormat` (`mp3` default / `wav` / `aac`). With both
188
- * sources each is split on its own. At most 200 parts per input. Poll
189
- * `jobs.get(jobId)`; the parts are on the completed job's `output_data` —
190
- * see {@link SplitMediaOutput}.
188
+ * sources each is split on its own. At most 200 parts per input. With
189
+ * `mode: "shots"` a video is instead cut into one part per shot, at the cuts
190
+ * measured from the file (no part length; a video only, up to 30 minutes and
191
+ * 200 shots; each part keeps the video's own sound). Poll `jobs.get(jobId)`;
192
+ * the parts are on the completed job's `output_data` — see
193
+ * {@link SplitMediaOutput}.
191
194
  */
192
195
  splitMedia(input) {
193
196
  return this.client.request("POST", "/v1/split-media", { body: input });
194
197
  }
198
+ /**
199
+ * Where every shot of a video starts and ends (`POST /v1/shot-cuts`),
200
+ * measured from the file: every decoded frame is scored and each hard cut
201
+ * starts a new shot (dissolves and fades are not cuts). Videos up to 30
202
+ * minutes. `frames: true` also saves each shot's exact first and last frame
203
+ * as PNG images (up to 1,000 shots; same price). Poll `jobs.get(jobId)`; the
204
+ * completed job's `output_data.json` is a {@link ShotCutsResult} — see
205
+ * {@link ShotCutsOutput}.
206
+ */
207
+ shotCuts(input) {
208
+ return this.client.request("POST", "/v1/shot-cuts", { body: input });
209
+ }
195
210
  /**
196
211
  * Turn one still image + one audio track into an MP4
197
212
  * (`POST /v1/still-to-video`) — locally rendered (FFmpeg), no AI model,
@@ -343,5 +358,5 @@ var MediaResource = class {
343
358
  };
344
359
 
345
360
  export { MediaResource };
346
- //# sourceMappingURL=chunk-72ZJ72WF.js.map
347
- //# sourceMappingURL=chunk-72ZJ72WF.js.map
361
+ //# sourceMappingURL=chunk-3HL3IHVE.js.map
362
+ //# sourceMappingURL=chunk-3HL3IHVE.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/resources/media.ts"],"names":[],"mappings":";;;AAsGO,IAAM,gBAAN,MAAoB;AAAA,EACzB,YAAoB,MAAA,EAA0B;AAA1B,IAAA,IAAA,CAAA,MAAA,GAAA,MAAA;AAAA,EAA2B;AAAA,EAA3B,MAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBpB,cAAc,KAAA,EAQsB;AAClC,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAgC,MAAA,EAAQ,sBAAsB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAClG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAcA,OAAO,qBAAA,CACL,UAAA,EACA,IAAA,GAAiC,EAAC,EACsB;AACxD,IAAA,MAAM,GAAA,GAAM,GAAG,IAAA,CAAK,MAAA,CAAO,OAAO,CAAA,4BAAA,EAA+B,kBAAA,CAAmB,UAAU,CAAC,CAAA,CAAA;AAC/F,IAAA,MAAM,KAAA,GAAQ,MAAM,IAAA,CAAK,MAAA,CAAO,KAAK,QAAA,EAAS;AAC9C,IAAA,MAAM,GAAA,GAAM,MAAM,IAAA,CAAK,MAAA,CAAO,MAAM,GAAA,EAAK;AAAA,MACvC,OAAA,EAAS,QAAQ,EAAE,aAAA,EAAe,UAAU,KAAK,CAAA,CAAA,KAAO,EAAC;AAAA,MACzD,QAAQ,IAAA,CAAK;AAAA,KACd,CAAA;AAID,IAAA,OAAO,aAAA,CAAqC,GAAA,EAAK,EAAE,KAAA,EAAO,mBAAmB,CAAA;AAAA,EAC/E;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,KAAA,EAAqH;AACjI,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,uBAAuB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC9F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkBA,aAAa,KAAA,EAYkB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,qBAAqB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC5F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAeA,aAAa,KAAA,EAiCkB;AAE7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,qBAAqB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC5F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,wBAAwB,KAAA,EAMoC;AAC1D,IAAA,OAAO,KAAK,MAAA,CAAO,OAAA;AAAA,MACjB,MAAA;AAAA,MACA,qCAAA;AAAA,MACA,EAAE,MAAM,KAAA;AAAM,KAChB;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,UAAU,KAAA,EAUqB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,kBAAkB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACzF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAuCA,YAAY,KAAA,EAAqD;AAC/D,IAAA,MAAM,EAAE,cAAA,EAAgB,kBAAA,EAAoB,GAAG,MAAK,GAAI,KAAA;AAGxD,IAAA,MAAM,IAAA,GAAgC;AAAA,MACpC,GAAG,IAAA;AAAA,MACH,GAAI,cAAA,KAAmB,MAAA,GAAY,EAAE,eAAA,EAAiB,cAAA,KAAmB,EAAC;AAAA,MAC1E,GAAI,kBAAA,KAAuB,MAAA,GAAY,EAAE,mBAAA,EAAqB,kBAAA,KAAuB;AAAC,KACxF;AACA,IAAA,OAAO,KAAK,MAAA,CAAO,OAAA,CAA2B,QAAQ,kBAAA,EAAoB,EAAE,MAAM,CAAA;AAAA,EACpF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,UAAU,KAAA,EAMqB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,kBAAkB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACzF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBA,WAAW,KAAA,EAAoD;AAC7D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,mBAAmB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC1F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,SAAS,KAAA,EAAkD;AACzD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,iBAAiB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACxF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,aAAa,KAAA,EAUkB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,sBAAsB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC7F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,UAAU,KAAA,EAcqB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,iBAAiB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACxF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkBA,aAAa,KAAA,EAAwD;AACnE,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,qBAAqB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC5F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EA+CA,UAAU,KAAA,EAAqD;AAC7D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,kBAAkB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACzF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAsBA,cAAc,KAAA,EAAyD;AACrE,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,sBAAsB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC7F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,KAAA,EAAgD;AAC5D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAuB,MAAA,EAAQ,sBAAsB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACzF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,OAAA,CACE,KAAA,EACA,OAAA,GAAwD,EAAC,EAClB;AACvC,IAAA,OAAO,IAAA,CAAK,MAAA,CAAO,OAAA,CAAsC,MAAA,EAAQ,mBAAA,EAAqB,EAAE,IAAA,EAAM,KAAA,EAAO,GAAG,OAAA,EAAS,CAAA;AAAA,EACnH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAaA,QAAQ,KAAA,EAA+D;AACrE,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAsC,MAAA,EAAQ,qBAAqB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACvG;AACF","file":"chunk-3HL3IHVE.js","sourcesContent":["import type { NodaroClientCore } from \"../core.js\"\nimport { readSseStream } from \"../sse.js\"\nimport type { OverlayAnchor, VideoMaskRequest, MaskCompositeRequest } from \"@nodaro/shared/media\"\nimport type { CaptionStyle, CaptionLookId, SupportedFontName } from \"@nodaro/shared/captions\"\nimport type { TranscribeLane } from \"@nodaro/shared/models\"\nimport type { VideoOverlayRequest } from \"@nodaro/shared/video\"\n\n/** One word-timed caption entry (one per WORD for the kinetic styles). */\nexport interface CaptionEntry {\n text: string\n startMs: number\n endMs: number\n timestampMs?: number | null\n confidence?: number | null\n}\n\n/** Look + motion levers — see {@link MediaResource.addCaptions}. `look` is a\n * named preset the explicit levers override; an unset look renders as `outline`\n * (the default). The STYLING levers (`look`, `fontFamily`, `fontWeight`,\n * `strokeColor`/`strokeWidth`, `uppercase`, `positionY`, `maxWordsPerLine`) now\n * apply to the static `subtitle` style too — a subtitle carrying any of them\n * renders via Remotion. Only `highlightColor` (no spoken-word cursor to colour)\n * and `animate` (no motion to freeze) stay kinetic-only and are rejected (400)\n * on `subtitle`. */\nexport interface CaptionLookInput {\n look?: CaptionLookId\n fontFamily?: SupportedFontName\n fontWeight?: number\n strokeColor?: string\n strokeWidth?: number\n highlightColor?: string\n uppercase?: boolean\n positionY?: number\n /** Cap on how many words one caption LINE (or `tiktok-words` page) may hold,\n * 1-20, on TOP of the frame-width budget, sentence ends and pauses that\n * already close a line — 1-2 gives the punchy CapCut read, unset fits the\n * width. Applies to `word-highlight` / `karaoke` / `bouncy` / `tiktok-words`\n * and to a Remotion-rendered `subtitle`; inert on `word-pop` (always one\n * word). */\n maxWordsPerLine?: number\n /** Kinetic styles only. `false` freezes the per-word MOTION — the grouping,\n * line-holding and spoken-word highlight stay, only the movement stops;\n * default `true`. Rejected (400) on the static `subtitle` style. */\n animate?: boolean\n}\n\n/** One caption SEGMENT: a time range with optional style/look overrides and its\n * own words (`text`/`captions`, else the shared transcript filtered to range).\n * A segmented render is entirely Remotion, so any `style` is valid here. */\nexport interface CaptionSegmentInput extends CaptionLookInput {\n startMs: number\n endMs: number\n style?: CaptionStyle\n position?: \"bottom\" | \"top\" | \"center\"\n fontSize?: number\n color?: string\n backgroundColor?: string\n text?: string\n captions?: CaptionEntry[]\n}\n\n/** {@link MediaResource.addCaptions} input. */\nexport interface AddCaptionsInput extends CaptionLookInput {\n videoUrl: string\n text?: string\n captions?: CaptionEntry[]\n /** Transcribe the video's audio when no text/captions are given (default true). */\n autoTranscribe?: boolean\n transcribeProvider?: TranscribeLane\n style?: CaptionStyle\n position?: \"bottom\" | \"top\" | \"center\"\n fontSize?: number\n color?: string\n backgroundColor?: string\n /** Apply different treatments to time ranges in one call (non-overlapping). */\n segments?: CaptionSegmentInput[]\n}\n\n/**\n * Media ingestion + trimming — the source-preparation steps a Voice Changer Pro\n * flow (or any pipeline) needs before it has a clip to work on: pull a social\n * video into storage, copy a remote URL into storage, trim a video/audio to a\n * range, and probe a video's metadata. Each generation-style op returns a job id\n * to poll (`jobs.get(jobId)`); `videoMetadata` is a direct read.\n */\n/**\n * Where one overlay layer should go, in {@link MediaResource.imageOverlay}'s\n * percent units — the answer from {@link MediaResource.suggestOverlayPlacement}.\n */\nexport interface OverlayPlacement {\n /** The point on the base image the layer attaches to. */\n anchor: OverlayAnchor\n /** Offset from the anchor in % of the base width (negative on a right anchor = inward). */\n x: number\n /** Offset from the anchor in % of the base height (negative on a bottom anchor = inward). */\n y: number\n /** The layer's width in % of the base width. */\n width: number\n /** One sentence, plain language, on why this spot. */\n reason: string\n}\n\nexport class MediaResource {\n constructor(private client: NodaroClientCore) {}\n\n /**\n * Download the video behind a link (the video's page, or a direct video file\n * link) into your storage (`POST /v1/download-video`). `maxHeight` caps the resolution\n * (default \"best\"); `sectionStartSec` + `sectionEndSec` (both-or-neither) fetch\n * ONLY that time range instead of the whole video. A download that arrives\n * with no audio stream FAILS by default (it is usually a degraded source\n * response, and is retried through other routes first); pass\n * `requireAudio: false` to accept a clip that really has no sound. Returns a\n * `downloadId`; progress streams from\n * `GET /v1/download-video/progress/:downloadId` (server-sent events) and the\n * finished file lands in your library. A section is fetched with ±3 s of\n * padding for the caller to trim; `exactSection: true` cuts it exactly at\n * `sectionStartSec`–`sectionEndSec` instead. `confirmationId` is the content\n * rights confirmation the import is made under (`contentRights.confirm`):\n * checked and consumed before the download starts, written on the stored\n * file; a download without one is refused (`403 rights_confirmation_required`)\n * once the deployment requires it, so a client always sends one.\n */\n downloadVideo(input: {\n url: string\n maxHeight?: number\n sectionStartSec?: number\n sectionEndSec?: number\n exactSection?: boolean\n requireAudio?: boolean\n confirmationId?: string\n }): Promise<{ downloadId: string }> {\n return this.client.request<{ downloadId: string }>(\"POST\", \"/v1/download-video\", { body: input })\n }\n\n /**\n * Stream the live progress of a {@link MediaResource.downloadVideo} import\n * (`GET /v1/download-video/progress/:downloadId`, server-sent events) as an\n * async iterable. Yields a {@link DownloadVideoProgress} roughly every 500ms\n * until the download reaches `completed` (its event carries the stored\n * `videoUrl`) or `failed` (its event carries `error`), then ends. The progress\n * state expires server-side shortly after the download starts existing, so\n * start iterating promptly after `downloadVideo` returns.\n *\n * No request timeout is applied (a large import legitimately takes minutes) —\n * pass an `AbortSignal` to cancel from the caller.\n */\n async *downloadVideoProgress(\n downloadId: string,\n opts: { signal?: AbortSignal } = {},\n ): AsyncGenerator<DownloadVideoProgress, void, undefined> {\n const url = `${this.client.baseUrl}/v1/download-video/progress/${encodeURIComponent(downloadId)}`\n const token = await this.client.auth.getToken()\n const res = await this.client.fetch(url, {\n headers: token ? { Authorization: `Bearer ${token}` } : {},\n signal: opts.signal,\n })\n // The shared reader owns the non-OK throw, the empty-body guard, the frame\n // parse and the cancel-on-early-exit — this route emits the same\n // `data: <json>` frames every other stream here does.\n yield* readSseStream<DownloadVideoProgress>(res, { label: \"progress stream\" })\n }\n\n /**\n * Copy an external media URL into your Nodaro storage (`POST /v1/save-to-storage`)\n * — a server-side fetch, so nothing round-trips through the client. Poll\n * `jobs.get(jobId)`.\n */\n saveToStorage(input: { mediaUrl: string; filename?: string; mediaType?: \"image\" | \"video\" | \"audio\" }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/save-to-storage\", { body: input })\n }\n\n /**\n * Composite 2–30 images into ONE large 2K/4K collage (`POST /v1/image-collage`).\n * `layout` `\"smart\"` (default — justified rows at each image's exact aspect\n * ratio, no cropping; the output height floats) or `\"grid\"` (uniform,\n * letterboxed cells). `imageSizes` is index-aligned with `imageUrls` and\n * gives per-image RELATIVE size hints for the smart layout: `0` auto\n * (\"don't care\", default), `1` big (~2× linear vs medium), `2` medium,\n * `3` small (~½ linear). All-equal hints change nothing; grid ignores them.\n * For storyboards, set `numbered` to stamp a 1-based sequence badge at each\n * image's corner (top-left by default; `badgePosition: \"top-right\"` moves it;\n * in `imageUrls` order) and pass `imageLabels`\n * (index-aligned with `imageUrls`; `null`/`\"\"`/omitted = no caption for that\n * image) to caption images after the number, e.g. `3 · Close-up`. Badges are\n * an overlay only — they never change the layout, the output size or the\n * credit cost. Poll `jobs.get(jobId)` for the finished image.\n */\n imageCollage(input: {\n imageUrls: string[]\n imageSizes?: Array<0 | 1 | 2 | 3>\n numbered?: boolean\n imageLabels?: Array<string | null>\n /** Corner the badges sit in — `\"top-left\"` (default, storyboard convention) or `\"top-right\"`. */\n badgePosition?: \"top-left\" | \"top-right\"\n layout?: \"smart\" | \"grid\"\n resolution?: \"2K\" | \"4K\"\n aspectRatio?: string\n gap?: number\n backgroundColor?: string\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/image-collage\", { body: input })\n }\n\n /**\n * Place 1–12 layers — pictures, real text, QR codes, shapes — on a base image,\n * pixel-exactly (`POST /v1/image-overlay`) — local, deterministic, no AI.\n * Every layer position/size is a PERCENT of the base image: `anchor` (nine\n * positions, default `\"center\"`), `x`/`y` (offset from the anchor in % of\n * the base width/height; negative on a right/bottom anchor moves inward),\n * `width` (% of the base width, default 25 — height follows the layer's\n * aspect unless `height` is set). Plus `opacity` (0..1), `rotation`\n * (degrees), `blend` (`\"over\"` | `\"multiply\"` | `\"screen\"`), `fit`,\n * `shadow` and `roundedCorners`. The output keeps the base's pixel size\n * unless `canvas` is set. Flat credit cost. Poll `jobs.get(jobId)` for the\n * finished image.\n */\n imageOverlay(input: {\n imageUrl: string\n layers: Array<{\n imageUrl: string\n anchor?: OverlayAnchor\n x?: number\n y?: number\n width?: number\n height?: number\n opacity?: number\n rotation?: number\n blend?: \"over\" | \"multiply\" | \"screen\"\n fit?: \"contain\" | \"cover\" | \"stretch\"\n shadow?: { blur: number; offsetX: number; offsetY: number; color: string; opacity: number }\n roundedCorners?: number\n zIndex?: number\n /** \"image\" (default, needs imageUrl) | \"text\" | \"qr\" | \"shape\" — see @nodaro/shared image-overlay-layers. */\n kind?: \"image\" | \"text\" | \"qr\" | \"shape\"\n text?: Record<string, unknown>\n qr?: Record<string, unknown>\n shape?: Record<string, unknown>\n effects?: Record<string, unknown>\n }>\n canvas?: { width: number; height: number; backgroundColor?: string }\n baseFit?: \"contain\" | \"cover\"\n outputFormat?: \"png\" | \"jpg\" | \"webp\"\n /** Extra platform renders (see OVERLAY_PLATFORMS in @nodaro/shared); returned as output `variants[]`. */\n variants?: string[]\n /** The mask the job also emits as output `maskUrl` (white = may change). Default \"around\". */\n maskMode?: \"none\" | \"layers\" | \"around\" | \"outside\"\n maskSpread?: number\n /** Fills every QR layer whose `qr.fromInput` is true — the node's QR link handle. */\n qrText?: string\n }): Promise<{ jobId: string }> {\n // The finished job's output: { imageUrl, width, height, maskUrl?, variants?: [{ id, label, width, height, url }] }.\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/image-overlay\", { body: input })\n }\n\n /**\n * Ask a vision model WHERE one overlay layer should sit on a base image\n * (`POST /v1/image-overlay/suggest-placement`) — it reads the picture and\n * keeps the element off the faces, the subject and the busiest texture.\n * The answer comes back in {@link MediaResource.imageOverlay}'s own percent\n * units — `anchor`, `x`/`y` offsets, `width` — so it drops straight onto a\n * layer (`const { reason, ...box } = placement`), plus a one-sentence\n * `reason` you can show a user. Nothing is composited here: apply the\n * placement yourself.\n *\n * `layerAspect` is the element's width / height (1 = square, the default) so\n * the proposed box stays in proportion; `intent` says what the element is\n * (\"a logo\", \"a price badge\"); `safeArea` is the always-visible region as\n * fractions of the canvas (a platform preset's safe area), which the\n * placement is kept inside. Unlike the other media calls this one answers\n * synchronously — there is nothing to poll; `jobId` is the billing record\n * (one image-to-text call).\n */\n suggestOverlayPlacement(input: {\n imageUrl: string\n layerAspect?: number\n intent?: string\n safeArea?: { x: number; y: number; w: number; h: number }\n llmModel?: string\n }): Promise<{ jobId: string; placement: OverlayPlacement }> {\n return this.client.request<{ jobId: string; placement: OverlayPlacement }>(\n \"POST\",\n \"/v1/image-overlay/suggest-placement\",\n { body: input },\n )\n }\n\n /**\n * Trim a video to a range (`POST /v1/trim-video`). Give the range in whichever\n * unit fits: `startTime`/`endTime` seconds, `trim*Frames`, `trim*Seconds`, or\n * `keepFirst`/`keepLastSeconds`. Poll `jobs.get(jobId)`.\n */\n trimVideo(input: {\n videoUrl: string\n startTime?: number\n endTime?: number\n trimStartFrames?: number\n trimEndFrames?: number\n trimStartSeconds?: number\n trimEndSeconds?: number\n keepFirstSeconds?: number\n keepLastSeconds?: number\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/trim-video\", { body: input })\n }\n\n /**\n * Burn captions into a video (`POST /v1/add-captions`). Give the words as\n * `text`, word-timed `captions[]`, or let it transcribe (`autoTranscribe`,\n * the default when neither is set).\n *\n * `style: \"subtitle\"` renders statically (FFmpeg) UNLESS it carries a styling\n * lever, in which case it — like the KINETIC styles (`word-highlight` /\n * `karaoke` / `tiktok-words` / `word-pop` / `bouncy`) — renders via Remotion.\n * `look` picks a preset — `outline` (Montserrat 900, UPPERCASE, black outline,\n * yellow spoken word — the TikTok/Reels read) or `clean`; an UNSET look renders\n * as `outline`. The explicit STYLING levers (`fontFamily`, `fontWeight`,\n * `strokeColor`/`strokeWidth`, `uppercase`, `positionY`, `maxWordsPerLine`)\n * override individual fields of it and now apply to `subtitle` too. Only\n * `highlightColor` (the spoken-word cursor) and `animate` stay kinetic-only\n * and are REJECTED on `subtitle`.\n *\n * `animate` (default true) freezes the per-word MOTION on the kinetic styles\n * when set to false — the grouping, line-holding and spoken-word highlight\n * stay; only the movement stops.\n *\n * `maxWordsPerLine` (1-20) caps how many words a line — or a `tiktok-words`\n * page — may hold, on top of the width budget / sentence ends / pauses that\n * already close one; it is inert on `word-pop`, which is always one word.\n * Like every lever here it exists top-level AND per segment, and a segment\n * that does not name its own inherits the top-level value.\n *\n * `segments[]` applies DIFFERENT treatments to non-overlapping time ranges in\n * one call (e.g. a large top intro, then a small bottom body); a segment that\n * names its own `look` starts fresh from that preset and does not inherit the\n * top-level explicit levers. Poll `jobs.get(jobId)`.\n *\n * `word-highlight` shows ONE held line at a time and a word's\n * `startMs`/`endMs` is its SPOKEN window (what times the highlight, not how\n * long the text is on screen) — so `captions[]` can be handed over verbatim\n * from an `audio.transcribe()` job's `output_data.words`, with\n * `autoTranscribe: false`.\n */\n addCaptions(input: AddCaptionsInput): Promise<{ jobId: string }> {\n const { autoTranscribe, transcribeProvider, ...rest } = input\n // The route body is camelCase except these two flags (worker snake_case);\n // omit each when unset so an absent flag stays absent on the wire.\n const body: Record<string, unknown> = {\n ...rest,\n ...(autoTranscribe !== undefined ? { auto_transcribe: autoTranscribe } : {}),\n ...(transcribeProvider !== undefined ? { transcribe_provider: transcribeProvider } : {}),\n }\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/add-captions\", { body })\n }\n\n /**\n * Trim (and extract) audio from a video or audio source\n * (`POST /v1/trim-audio`) to `[startTime, endTime]` seconds, in `audioFormat`\n * (`mp3` default / `wav` / `aac`). Poll `jobs.get(jobId)`.\n */\n trimAudio(input: {\n videoUrl?: string\n audioUrl?: string\n audioFormat?: \"mp3\" | \"wav\" | \"aac\"\n startTime?: number\n endTime?: number\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/trim-audio\", { body: input })\n }\n\n /**\n * Cut a video and/or an audio file into consecutive parts\n * (`POST /v1/split-media`). Every part is at most `chunkDuration` seconds and\n * at least `minPartDuration` (default 4, never above `chunkDuration`) — a\n * short leftover is evened out with the parts before it. Video is cut on\n * whole frames with each part's sound matching its picture; audio parts are\n * sample-exact, in `audioFormat` (`mp3` default / `wav` / `aac`). With both\n * sources each is split on its own. At most 200 parts per input. With\n * `mode: \"shots\"` a video is instead cut into one part per shot, at the cuts\n * measured from the file (no part length; a video only, up to 30 minutes and\n * 200 shots; each part keeps the video's own sound). Poll `jobs.get(jobId)`;\n * the parts are on the completed job's `output_data` — see\n * {@link SplitMediaOutput}.\n */\n splitMedia(input: SplitMediaInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/split-media\", { body: input })\n }\n\n /**\n * Where every shot of a video starts and ends (`POST /v1/shot-cuts`),\n * measured from the file: every decoded frame is scored and each hard cut\n * starts a new shot (dissolves and fades are not cuts). Videos up to 30\n * minutes. `frames: true` also saves each shot's exact first and last frame\n * as PNG images (up to 1,000 shots; same price). Poll `jobs.get(jobId)`; the\n * completed job's `output_data.json` is a {@link ShotCutsResult} — see\n * {@link ShotCutsOutput}.\n */\n shotCuts(input: ShotCutsInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/shot-cuts\", { body: input })\n }\n\n /**\n * Turn one still image + one audio track into an MP4\n * (`POST /v1/still-to-video`) — locally rendered (FFmpeg), no AI model,\n * zero credits. The output duration IS the audio's duration; there is no\n * duration field. Optional `motion` animates the still (zoom / pan /\n * ken-burns) at `intensity` 1–10. `fit: \"contain\"` letterboxes with\n * `padColor` instead of cropping. Poll `jobs.get(jobId)`.\n */\n stillToVideo(input: {\n imageUrl: string\n audioUrl: string\n motion?: \"none\" | \"zoom-in\" | \"zoom-out\" | \"pan-left\" | \"pan-right\" | \"ken-burns\"\n intensity?: number\n resolution?: \"720p\" | \"1080p\" | \"4K\"\n aspectRatio?: \"16:9\" | \"9:16\" | \"1:1\" | \"4:3\"\n fps?: 24 | 30\n fit?: \"cover\" | \"contain\"\n padColor?: string\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/still-to-video\", { body: input })\n }\n\n /**\n * Turn 2–100 images + one optional audio track into an MP4 slideshow\n * (`POST /v1/slideshow`) — locally rendered (FFmpeg), zero credits. With\n * audio, the output duration IS the audio's duration (equal split unless\n * `imageDurations` pins rows — null entries = auto; mismatched pinned sums\n * scale proportionally and the factor is disclosed in the job output).\n * Without audio: N × `perImageDuration`, silent output. Transitions\n * consume the outgoing slide, so totals stay exact. For a single image use\n * `stillToVideo`. Poll `jobs.get(jobId)`.\n */\n slideshow(input: {\n imageUrls: string[]\n audioUrl?: string\n imageDurations?: Array<number | null>\n perImageDuration?: number\n transition?: string\n transitionDuration?: number\n motion?: \"none\" | \"zoom-in\" | \"zoom-out\" | \"ken-burns\" | \"alternate\"\n intensity?: number\n resolution?: \"720p\" | \"1080p\" | \"4K\"\n aspectRatio?: \"16:9\" | \"9:16\" | \"1:1\" | \"4:3\"\n fps?: 24 | 30\n fit?: \"cover\" | \"contain\"\n padColor?: string\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/slideshow\", { body: input })\n }\n\n /**\n * Place 1–20 timed image layers over a video (`POST /v1/video-overlay`) —\n * rendered locally (FFmpeg) in one pass, no AI; the base audio is kept\n * untouched. Each layer has an `imageUrl`, a `start` and an optional `end`\n * in seconds (no `end` = to the end of the video) and a placement: a\n * `preset` (`\"card\"`, `\"corner-badge\"` with a `corner`, `\"full-frame\"`) or\n * an explicit box in {@link MediaResource.imageOverlay}'s percent vocabulary\n * (`anchor`, `x`, `y`, `width`, `height`, `fit`) — an explicit box field\n * overrides the preset, and a layer with neither is a corner badge\n * (bottom-right, or the `corner` it names). `outputAspect` renders onto a\n * 16:9 / 9:16 / 1:1 / 4:5 canvas\n * (`baseFit` cover by default; `backgroundColor` pads `contain`). Poll\n * `jobs.get(jobId)`: the output carries `videoUrl`, `thumbnailUrl`, `width`,\n * `height`, `durationSec` and `warnings[]` (a layer clipped or skipped at the\n * video's end, an animated image's first frame, re-encoded audio).\n */\n videoOverlay(input: VideoOverlayRequest): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/video-overlay\", { body: input })\n }\n\n /**\n * Track a subject described in plain text through a video and get a\n * black-and-white MASK VIDEO back (`POST /v1/video-mask`, SAM 3.1 by default,\n * or `provider: \"sam-3\"` at half the price) — white = the subject. Clips up\n * to 5 minutes at 30 fps (a faster clip counts\n * proportionally longer), priced by the clip's length. Poll\n * `jobs.get(jobId)`: the output carries `videoUrl` (the mask), `width`,\n * `height`, `frames`, `coverage` (the fraction of pixels selected, rounded\n * to 4 decimals) and `kept` (any region left on any frame: false only when\n * the size / zone filters left an empty mask — read it rather than\n * `coverage`, which a few pixels round to 0), and `recipe` (the version of\n * the masking algorithm that made it, `VIDEO_MASK_RECIPE`: it changes\n * whenever the same clip and settings would give a different mask). Feed the mask to\n * {@link MediaResource.maskComposite}.\n *\n * `holdLastMask: true` keeps the last mask on frames where the subject is\n * lost (a mask covering under 0.002 % of the frame, e.g. a face behind a\n * hand): each such frame after the first detection repeats the most recent\n * non-empty one, for at most 1 s of frames (`VIDEO_MASK_LIMITS.HOLD_MAX_SEC`;\n * longer gaps stay empty after that); the output then also carries `heldFrames`.\n * `minSubjectHeightPct` (0–100, default 0 = off) drops every connected\n * region whose bounding box is shorter than that % of the frame height\n * (extras' small faces behind the main subject); it runs before\n * `holdLastMask`, and the output then carries `droppedRegions`.\n * `maxSubjectHeightPct` (0–100, default 0 = off) is its mirror: it drops\n * regions TALLER than that % (keep only the small background faces). With\n * both set a region is kept when min ≤ height ≤ max; min > max is a 400.\n * `zone` (`{ leftPct, rightPct, topPct, bottomPct }`, each 0–49 % of the\n * frame, a margin left out = 0; default null = off) keeps only the regions\n * whose bounding-box centre lies inside the frame minus those margins (after\n * the size cuts, before `holdLastMask`). `perShot: true` finds the clip's\n * scene cuts first and segments each shot on its own: the subject is\n * detected again after every cut, `holdLastMask` never crosses a cut, and\n * each shot's mask is fitted to that shot's exact frame count; on a clip\n * with no cut it gives the same mask, fitted to the clip's exact frame\n * count. The output carries `shots` (1 when off). `redetectEverySec` (0 =\n * off, the default, or 1–10) also detects the subject again every that\n * many seconds — the clip (each shot, with `perShot`) is sent to the model\n * in pieces of at most that length, a last piece under 8 frames joining the\n * one before — so someone who walks in or turns to the camera mid-shot is\n * masked from the next boundary on. These boundaries are not cuts:\n * `holdLastMask` holds across them. Each piece is one more model request\n * (three run at a time), so it is slower; the price does not change. The\n * output's `chunks` counts the requests.\n */\n videoMask(input: VideoMaskRequest): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/video-mask\", { body: input })\n }\n\n /**\n * Place a foreground video over a background video through a mask video\n * (`POST /v1/mask-composite`; white = foreground) — local FFmpeg, no AI.\n * The result follows the foreground's size, frame rate and length; the\n * background is fitted (`backgroundFit`, default cover) and loops when\n * shorter; the mask holds its last frame when shorter. Optional\n * `invertMask`, `feather` (0–20 px), `grow` (−10–10 px), `opacity` (0–1,\n * default 1: how strongly the foreground covers the background inside the\n * mask) and `audioSource` (foreground / background / mix / none). Priced per started 5 seconds of\n * the result (the video it follows up to 300 s). Poll `jobs.get(jobId)` for `videoUrl`.\n *\n * `fill` picks what fills the masked area: `video` (default — the\n * foreground, required then), `pixelate` (the background itself in\n * blocks of `pixelSizePct` % of the frame width, 0.2–10, default 1.25) or\n * `blur` (the background blurred, sigma `blurStrengthPct` % of the frame\n * width, 0.1–5, default 2). Pixelate and blur take no\n * `foregroundUrl` (one sent is ignored): the result follows the background\n * and is priced on its length, and it always carries the background's sound\n * (`audioSource` is ignored).\n */\n maskComposite(input: MaskCompositeRequest): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/mask-composite\", { body: input })\n }\n\n /**\n * Probe a social video's metadata (`POST /v1/video-metadata`) — duration,\n * dimensions, title, live status — WITHOUT downloading it. A direct read, not a\n * job. Use it to decide whether to trim before importing.\n */\n videoMetadata(input: { url: string }): Promise<VideoMetadata> {\n return this.client.request<VideoMetadata>(\"POST\", \"/v1/video-metadata\", { body: input })\n }\n\n /**\n * Cut or crop a stored file (`POST /v1/media/process`) — synchronous and\n * free, the source-preparation sibling of the priced `trimVideo` node.\n * `trim` is `[startTime, endTime]` in seconds; `deleteSource: true` removes\n * the source object afterwards when it is yours and nothing else references\n * it (\"the cut replaces the original\"). Answers the stored file. The\n * platform may take up to 300 s on a long file: pass `timeoutMs` to wait\n * longer than the client's default.\n */\n process(\n input: MediaProcessInput,\n options: { timeoutMs?: number; signal?: AbortSignal } = {},\n ): Promise<{ data: MediaProcessResult }> {\n return this.client.request<{ data: MediaProcessResult }>(\"POST\", \"/v1/media/process\", { body: input, ...options })\n }\n\n /**\n * Throw away a file you just made and no longer need (`POST /v1/media/discard`):\n * a link download replaced before use, a cut superseded by a newer one, a\n * picture imported then cancelled. Not a general delete. The file goes only\n * when it is your own, was made by a link download (`downloadVideo`), a cut\n * (`process`) or a URL import, is under 24 hours old, and no job of yours\n * uses it (as an output or as an input such as a run's `videoUrl`);\n * otherwise it is left untouched and the call fails with 409\n * `not_discardable`. `discarded: false` means there was nothing of yours at\n * that address (already gone, or not yours).\n */\n discard(input: { url: string }): Promise<{ data: MediaDiscardResult }> {\n return this.client.request<{ data: MediaDiscardResult }>(\"POST\", \"/v1/media/discard\", { body: input })\n }\n}\n\n/**\n * One event from {@link MediaResource.downloadVideoProgress}. The stream ends\n * after a `completed` event (which carries the stored `videoUrl` + an optional\n * `thumbnailUrl`) or a `failed` event (which carries `error`).\n */\nexport interface DownloadVideoProgress {\n phase: \"downloading\" | \"processing\" | \"uploading\" | \"completed\" | \"failed\"\n /** Download percent (0–100). Section fetches report jumpy percents — display, don't sum. */\n percent: number\n /** The imported video's storage URL — set on the `completed` event. */\n videoUrl?: string\n /** Thumbnail storage URL — set on the `completed` event when one was captured. */\n thumbnailUrl?: string\n /** What went wrong — set on the `failed` event. */\n error?: string\n}\n\n/** Result of {@link MediaResource.videoMetadata}. Fields are best-effort — a probe may omit some. */\nexport interface VideoMetadata {\n durationSec?: number | null\n width?: number | null\n height?: number | null\n title?: string | null\n isLive?: boolean\n [key: string]: unknown\n}\n\n/** Body of {@link MediaResource.splitMedia} (`POST /v1/split-media`): parts of a set length, or one part per shot of a video. */\nexport type SplitMediaInput = SplitMediaByDurationInput | SplitMediaByShotsInput\n\n/** {@link SplitMediaInput} cutting parts of at most `chunkDuration` seconds. At least one of `videoUrl` / `audioUrl`. */\nexport interface SplitMediaByDurationInput {\n /** Video to split. */\n videoUrl?: string\n /** Audio file to split (split on its own when a video is given too). */\n audioUrl?: string\n /** The default: parts of a set length. */\n mode?: \"duration\"\n /** The longest a part may be, in seconds (at least 1). Every part is at most this long. */\n chunkDuration: number\n /** The shortest a part may be, in seconds (0 or more, default 4). Must not exceed `chunkDuration` (400 otherwise). */\n minPartDuration?: number\n /** Format of the audio parts. Default `mp3`. */\n audioFormat?: \"mp3\" | \"wav\" | \"aac\"\n}\n\n/** {@link SplitMediaInput} cutting a video into one part per shot, at its measured cuts. */\nexport interface SplitMediaByShotsInput {\n /** The video to split (up to 30 minutes). A separate audio file is refused (each part keeps the video's own sound). */\n videoUrl: string\n mode: \"shots\"\n}\n\n/** One part's shot in a shots-mode split: its place on the video, seconds to the millisecond. */\nexport interface SplitMediaShot {\n index: number\n startSec: number\n endSec: number\n durationSec: number\n}\n\n/** A completed split-media job's `output_data`. */\nexport interface SplitMediaOutput {\n /** The video parts, in order (when a video was split). */\n videoUrls?: string[]\n /** The audio parts, in order (when an audio file was split). */\n audioUrls?: string[]\n /** The number of parts — the larger of the two lists. */\n chunkCount: number\n /** Each part's exact length in seconds: the video parts' when a video was split, else the audio parts'. */\n partDurationsSec?: number[]\n /** The audio parts' lengths, only when BOTH a video and an audio file were split. */\n audioPartDurationsSec?: number[]\n /** Shots mode: the number of shots (= parts). */\n shotCount?: number\n /** Shots mode: each part's shot, in part order. */\n shots?: SplitMediaShot[]\n}\n\n/** Body of {@link MediaResource.shotCuts} (`POST /v1/shot-cuts`). */\nexport interface ShotCutsInput {\n /** The video to measure: MP4, MOV, MKV or WebM, up to 30 minutes. */\n videoUrl: string\n /** Also save each shot's first and last frame as images (up to 1,000 shots). Default false. */\n frames?: boolean\n}\n\n/** One shot of a {@link ShotCutsResult}: seconds from the video's first frame, to the millisecond. */\nexport interface ShotCutsShot {\n index: number\n startSec: number\n /** Where the next shot starts, or the video's end. */\n endSec: number\n durationSec: number\n /** With `frames`: the shot's first decoded frame (PNG). */\n firstFrameUrl?: string\n /** With `frames`: its last decoded frame — the one before the next cut, or the video's last. */\n lastFrameUrl?: string\n}\n\n/** What a Shot Cuts run measured (`output_data.json`). */\nexport interface ShotCutsResult {\n shotCount: number\n /** Where each new shot starts: ascending, never 0; `[]` for a single shot. */\n cutsSec: number[]\n /** The video's length. */\n durationSec: number\n shots: ShotCutsShot[]\n}\n\n/** A completed shot-cuts job's `output_data`. */\nexport interface ShotCutsOutput {\n json: ShotCutsResult\n /** With `frames`: each shot's first frame, in shot order. */\n firstFrames?: string[]\n /** With `frames`: each shot's last frame, in shot order. */\n lastFrames?: string[]\n}\n\nexport interface MediaProcessInput {\n sourceUrl: string\n type: \"video\" | \"audio\"\n trim?: { startTime: number; endTime: number }\n crop?: { x: number; y: number; width: number; height: number }\n format?: \"mp4\" | \"webm\" | \"mp3\" | \"wav\" | \"m4a\" | \"aac\"\n deleteSource?: boolean\n}\n\n/** Result of {@link MediaResource.discard}. */\nexport interface MediaDiscardResult {\n /** True when the file was deleted; false when there was nothing of yours at that address. */\n discarded: boolean\n}\n\nexport interface MediaProcessResult {\n url: string\n thumbnailUrl: string | null\n assetId: string | null\n sizeBytes: number\n mimeType: string\n metadata: Record<string, unknown>\n}\n"]}
@@ -13,8 +13,10 @@ var ContentRightsResource = class {
13
13
  }
14
14
  /**
15
15
  * `POST /v1/content-rights/confirm` → the confirmation for one link, under
16
- * the declaration version the person saw. `409 rights_confirmation_stale`
17
- * when the declaration changed since: read it again and ask again.
16
+ * the declaration version the person saw. Any `http(s)` link an import takes
17
+ * can be confirmed (a video's page, a file, a picture). `409
18
+ * rights_confirmation_stale` when the declaration changed since: read it
19
+ * again and ask again.
18
20
  */
19
21
  async confirm(input) {
20
22
  const res = await this.client.request("POST", "/v1/content-rights/confirm", { body: input });
@@ -49,5 +51,5 @@ var ContentRightsResource = class {
49
51
  };
50
52
 
51
53
  exports.ContentRightsResource = ContentRightsResource;
52
- //# sourceMappingURL=chunk-3OARKXWT.cjs.map
53
- //# sourceMappingURL=chunk-3OARKXWT.cjs.map
54
+ //# sourceMappingURL=chunk-4DCXJ77J.cjs.map
55
+ //# sourceMappingURL=chunk-4DCXJ77J.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/resources/content-rights.ts"],"names":[],"mappings":";;;AA8CO,IAAM,wBAAN,MAA4B;AAAA,EACjC,YAAoB,MAAA,EAA0B;AAA1B,IAAA,IAAA,CAAA,MAAA,GAAA,MAAA;AAAA,EAA2B;AAAA,EAA3B,MAAA;AAAA;AAAA,EAGpB,MAAM,WAAA,GAAiD;AACrD,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAA4C,OAAO,gCAAgC,CAAA;AACjH,IAAA,OAAO,GAAA,CAAI,IAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAM,QAAQ,KAAA,EAAwF;AACpG,IAAA,MAAM,GAAA,GAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAA6C,QAAQ,4BAAA,EAA8B,EAAE,IAAA,EAAM,KAAA,EAAO,CAAA;AAChI,IAAA,OAAO,GAAA,CAAI,IAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,MAAM,aAAA,GAA4D;AAChE,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAAqD,OAAO,kCAAkC,CAAA;AAC5H,IAAA,OAAO,GAAA,CAAI,IAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAM,UAAU,KAAA,EAA4E;AAC1F,IAAA,MAAM,GAAA,GAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAA8C,QAAQ,kCAAA,EAAoC,EAAE,IAAA,EAAM,KAAA,EAAO,CAAA;AACvI,IAAA,OAAO,GAAA,CAAI,IAAA;AAAA,EACb;AAAA;AAAA,EAGA,MAAM,mBAAA,GAAqD;AACzD,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAAwC,UAAU,kCAAkC,CAAA;AAClH,IAAA,OAAO,GAAA,CAAI,IAAA;AAAA,EACb;AACF","file":"chunk-4DCXJ77J.cjs","sourcesContent":["import type { NodaroClientCore } from \"../core.js\"\n\n/**\n * Content rights — the declaration a person confirms before a link is\n * imported, and the confirmation a download then carries. Mirrors\n * `/v1/content-rights/*` (`backend/src/routes/content-rights.ts`).\n *\n * The flow: show {@link ContentRightsResource.declaration}'s text with a\n * checkbox, call {@link ContentRightsResource.confirm} with the link once it\n * is ticked, and pass the answered `id` as `confirmationId` to\n * `media.downloadVideo`. One confirmation is good for one download of that\n * link, by the same account, within an hour. Whether a download WITHOUT a\n * confirmation is refused is the deployment's setting\n * (`content_rights_required`); a client always sends one.\n *\n * An import through an API key or an AI agent has nobody to tick a box: the\n * person authorizes those imports ONCE, in their own session, with\n * {@link ContentRightsResource.authorize} (the standing authorization), and\n * every such import is then recorded as a confirmation derived from it.\n * {@link ContentRightsResource.authorization} reads the current one and\n * {@link ContentRightsResource.revokeAuthorization} ends it.\n */\n\nexport interface ContentRightsDeclaration {\n /** The declaration's version; a confirmation names the version it agreed to. */\n version: number\n /** The text the person confirms, verbatim. */\n text: string\n}\n\nexport interface ContentRightsConfirmation {\n /** The id a download carries as `confirmationId`. */\n id: string\n declarationVersion: number\n /** After this the confirmation no longer opens a download (ISO 8601). */\n expiresAt: string\n}\n\nexport interface ContentRightsAuthorization {\n id: string\n /** The declaration version the person authorized; a newer declaration asks for a new authorization. */\n declarationVersion: number\n /** When it was given (ISO 8601). */\n createdAt: string\n}\n\nexport class ContentRightsResource {\n constructor(private client: NodaroClientCore) {}\n\n /** `GET /v1/content-rights/declaration` → the current text and its version. Public: needs no credential. */\n async declaration(): Promise<ContentRightsDeclaration> {\n const res = await this.client.request<{ data: ContentRightsDeclaration }>(\"GET\", \"/v1/content-rights/declaration\")\n return res.data\n }\n\n /**\n * `POST /v1/content-rights/confirm` → the confirmation for one link, under\n * the declaration version the person saw. Any `http(s)` link an import takes\n * can be confirmed (a video's page, a file, a picture). `409\n * rights_confirmation_stale` when the declaration changed since: read it\n * again and ask again.\n */\n async confirm(input: { url: string; declarationVersion: number }): Promise<ContentRightsConfirmation> {\n const res = await this.client.request<{ data: ContentRightsConfirmation }>(\"POST\", \"/v1/content-rights/confirm\", { body: input })\n return res.data\n }\n\n /**\n * `GET /v1/content-rights/authorization` → the account's standing\n * authorization for imports through its API keys and AI agents, or `null`\n * when there is none (never given, revoked, or given under an older\n * declaration than the current one).\n */\n async authorization(): Promise<ContentRightsAuthorization | null> {\n const res = await this.client.request<{ data: ContentRightsAuthorization | null }>(\"GET\", \"/v1/content-rights/authorization\")\n return res.data\n }\n\n /**\n * `POST /v1/content-rights/authorization` → records that the person\n * authorized imports through their keys and agents under the declaration\n * version they saw. The person's own session only: a key or an agent\n * answers `403 forbidden`. `409 rights_confirmation_stale` when the\n * declaration changed since: read it again and ask again.\n */\n async authorize(input: { declarationVersion: number }): Promise<ContentRightsAuthorization> {\n const res = await this.client.request<{ data: ContentRightsAuthorization }>(\"POST\", \"/v1/content-rights/authorization\", { body: input })\n return res.data\n }\n\n /** `DELETE /v1/content-rights/authorization` → revokes the standing authorization (the record stays). The person's own session only. */\n async revokeAuthorization(): Promise<{ revoked: boolean }> {\n const res = await this.client.request<{ data: { revoked: boolean } }>(\"DELETE\", \"/v1/content-rights/authorization\")\n return res.data\n }\n}\n"]}
@@ -194,5 +194,5 @@ var VoicesResource = class {
194
194
  };
195
195
 
196
196
  export { VoicesResource };
197
- //# sourceMappingURL=chunk-T7N7DFQJ.js.map
198
- //# sourceMappingURL=chunk-T7N7DFQJ.js.map
197
+ //# sourceMappingURL=chunk-CRIQLLR2.js.map
198
+ //# sourceMappingURL=chunk-CRIQLLR2.js.map
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/resources/voices.ts"],"names":[],"mappings":";AAoBO,IAAM,iBAAN,MAAqB;AAAA,EAC1B,YAAoB,MAAA,EAA0B;AAA1B,IAAA,IAAA,CAAA,MAAA,GAAA,MAAA;AAAA,EAA2B;AAAA,EAA3B,MAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMpB,MAAM,IAAA,GAAyB;AAC7B,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAA6B,OAAO,YAAY,CAAA;AAC9E,IAAA,OAAO,GAAA,CAAI,MAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,aAAA,CAAc,MAAA,GAA6B,EAAC,EAAkC;AAC5E,IAAA,MAAM,QAA+D,EAAC;AACtE,IAAA,KAAA,MAAW,CAAC,CAAA,EAAG,CAAC,KAAK,MAAA,CAAO,OAAA,CAAQ,MAAM,CAAA,EAAG;AAC3C,MAAA,IAAI,CAAA,KAAM,UAAa,CAAA,KAAM,IAAA,IAAQ,MAAM,EAAA,EAAI,KAAA,CAAM,CAAC,CAAA,GAAI,CAAA;AAAA,IAC5D;AACA,IAAA,OAAO,KAAK,MAAA,CAAO,OAAA,CAAQ,OAAO,oBAAA,EAAsB,EAAE,OAAO,CAAA;AAAA,EACnE;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,UAAA,GAAoC;AACxC,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAAuC,OAAO,kBAAkB,CAAA;AAC9F,IAAA,OAAO,GAAA,CAAI,WAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,YAAY,KAAA,EAAgE;AAC1E,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAoB,MAAA,EAAQ,6BAA6B,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC7F;AAAA;AAAA,EAGA,MAAM,YAAY,EAAA,EAA2B;AAC3C,IAAA,MAAM,IAAA,CAAK,OAAO,OAAA,CAAc,QAAA,EAAU,oBAAoB,kBAAA,CAAmB,EAAE,CAAC,CAAA,CAAE,CAAA;AAAA,EACxF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAaA,OAAO,KAAA,EAewB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,qBAAqB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC5F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoCA,OAAO,KAAA,EAAyD;AAC9D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,yBAAyB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAChG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBA,QAAQ,KAAA,EAAoD;AAC1D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,iCAAiC,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACxG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAeA,UAAU,KAAA,EAAmD;AAC3D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,gCAAgC,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACvG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAO,KAAA,EAAqD;AAC1D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,oBAAoB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC3F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,KAAA,EAAoD;AACxD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,mBAAmB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC1F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,IAAI,KAAA,EAAiD;AACnD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,eAAe,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACtF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,eAAe,KAAA,EAcgB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,wBAAwB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC/F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,oBAAoB,KAAA,EAOI;AACtB,IAAA,MAAM,IAAA,GAAO,IAAI,QAAA,EAAS;AAG1B,IAAA,IAAA,CAAK,MAAA,CAAO,MAAA,EAAQ,KAAA,CAAM,IAAI,CAAA;AAC9B,IAAA,MAAM,OACJ,KAAA,CAAM,IAAA,YAAgB,IAAA,GAClB,KAAA,CAAM,OACN,IAAI,IAAA,CAAK,CAAC,KAAA,CAAM,IAAgB,CAAA,EAAG,EAAE,MAAM,KAAA,CAAM,WAAA,IAAe,cAAc,CAAA;AACpF,IAAA,IAAA,CAAK,MAAA,CAAO,MAAA,EAAQ,IAAA,EAAM,KAAA,CAAM,YAAY,QAAQ,CAAA;AACpD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAoB,MAAA,EAAQ,oBAAoB,EAAE,IAAA,EAAM,MAAM,CAAA;AAAA,EACnF;AACF","file":"chunk-T7N7DFQJ.js","sourcesContent":["import type { Voice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from \"@nodaro/shared/pipeline\"\nimport type { AudioFxPreset } from \"@nodaro/shared/media\"\nimport type { DialogueProvider } from \"@nodaro/shared/models\"\nexport type { Voice, SharedVoice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from \"@nodaro/shared/pipeline\"\nexport type { AudioFxPreset } from \"@nodaro/shared/media\"\nimport type { NodaroClientCore } from \"../core.js\"\n\n/**\n * Re-export the shared voice types so SDK consumers don't have to add\n * `@nodaro/shared` as a second dependency just to type a `Voice` row, a\n * `VoiceClone`, or a `searchLibrary` call. Single source of truth lives in\n * `@nodaro/shared`.\n */\n/** Audio-FX preset union (reverb spaces / telephone / megaphone / echo / custom) — used by {@link VoiceChangerProInput.voiceFx}. */\n\n/**\n * Read access to ElevenLabs voices: the premade catalog plus the shared\n * community Voice Library (both public GETs, no body), and the signed-in\n * user's own pre-retirement voice clones (list / delete).\n */\nexport class VoicesResource {\n constructor(private client: NodaroClientCore) {}\n\n /**\n * List the premade ElevenLabs voices (`GET /v1/voices`). Falls back to a\n * curated set server-side when no ElevenLabs API key is configured.\n */\n async list(): Promise<Voice[]> {\n const res = await this.client.request<{ voices: Voice[] }>(\"GET\", \"/v1/voices\")\n return res.voices\n }\n\n /**\n * Search the shared/community Voice Library (`GET /v1/voices/library`). All\n * params are optional and forwarded as a querystring; `undefined` / `null` /\n * empty-string values are omitted so the server defaults apply. `hasMore`\n * drives \"load more\" pagination.\n */\n searchLibrary(params: VoiceLibraryParams = {}): Promise<VoiceLibraryResponse> {\n const query: Record<string, string | number | boolean | undefined> = {}\n for (const [k, v] of Object.entries(params)) {\n if (v !== undefined && v !== null && v !== \"\") query[k] = v as string | number | boolean\n }\n return this.client.request(\"GET\", \"/v1/voices/library\", { query })\n }\n\n /**\n * List the signed-in user's voice clones (`GET /v1/voice-clones`). The route\n * wraps the rows in `{ voiceClones }`; we unwrap to the bare array.\n */\n async listClones(): Promise<VoiceClone[]> {\n const res = await this.client.request<{ voiceClones: VoiceClone[] }>(\"GET\", \"/v1/voice-clones\")\n return res.voiceClones\n }\n\n /**\n * @deprecated Voice cloning is no longer offered on Nodaro (retired\n * 2026-09-15). The route answers `410 voice_cloning_retired` and this call\n * rejects with that error. Clones created before the retirement are still\n * listed by {@link VoicesResource.listClones} and still work as voice ids.\n * For a new custom voice use {@link VoicesResource.design}.\n */\n createClone(input: { name: string; audioUrl: string }): Promise<VoiceClone> {\n return this.client.request<VoiceClone>(\"POST\", \"/v1/voice-clones/from-url\", { body: input })\n }\n\n /** Delete one of the user's voice clones (`DELETE /v1/voice-clones/:id`). */\n async deleteClone(id: string): Promise<void> {\n await this.client.request<void>(\"DELETE\", `/v1/voice-clones/${encodeURIComponent(id)}`)\n }\n\n /**\n * Replace the voice in a recording — or in a whole talking video — with a\n * different voice (`POST /v1/voice-changer`). Pass `audioUrl` to revoice\n * audio→audio, or `videoUrl` to revoice an entire clip (the server demuxes\n * the audio, runs speech-to-speech, and remuxes onto the original video,\n * returning the video plus the new audio track). Exactly one of `audioUrl` /\n * `videoUrl` is required; when both are sent, video wins. `removeBackgroundNoise`\n * off keeps the music/SFX bed under the new voice; on yields a clean voice-only\n * result. Costs credits and runs async — poll `jobs.get(jobId)` for the result\n * (`output_data.videoUrl` + `output_data.audioUrl` in video mode).\n */\n change(input: {\n voiceId: string\n audioUrl?: string\n videoUrl?: string\n /** Speech-to-speech model id. Defaults to the server-configured default when omitted. */\n model?: string\n stability?: number\n similarityBoost?: number\n /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency/stability. */\n style?: number\n /** ElevenLabs speaker boost — sharpens fidelity to the target speaker (small latency cost). */\n useSpeakerBoost?: boolean\n /** Deterministic speech-to-speech seed (integer 0–4294967295) for reproducible output. Omit for random. */\n seed?: number\n removeBackgroundNoise?: boolean\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer\", { body: input })\n }\n\n /**\n * Recast each detected speaker in a multi-speaker recording to a different\n * voice (`POST /v1/voice-changer-pro`). `orderedVoices` maps speaker positions to\n * voices in detection order — speaker 0 → `orderedVoices[0]`, speaker 1 →\n * `orderedVoices[1]`, etc. Speakers beyond the end of `orderedVoices` keep\n * their original voice. Each entry is EITHER a bare voice id (premade name or\n * ElevenLabs UUID) OR a {@link VoiceChangerProVoice} object carrying per-voice\n * ElevenLabs speech-to-speech settings (stability / similarityBoost / style /\n * useSpeakerBoost / `seed`) plus a loudness `volumeMode` (and a manual\n * `volume`) — OR `null`, meaning keep this speaker's original voice (requires\n * a platform running cloud-plugins with keep-slot support). At least one\n * entry must be non-null. A per-voice `seed` makes that speaker's recast\n * reproducible.\n *\n * Pass `audioUrl` for audio-only recast or `videoUrl` to recast the audio\n * track of a video clip (the server demuxes, recasts, and remuxes).\n *\n * Voice and music are ALWAYS separated first — before recasting, the source\n * is split into an isolated vocal stem and a music/SFX stem.\n * `preserveBackground` (default `true`) only controls whether that\n * music/instrumental stem is mixed back under the new voices; set it `false`\n * for a clean voice-only result. `separationQuality` selects the quality of\n * the voice/music separation: `\"fast\"` (default, quicker — preserves more of\n * the voice) or `\"best\"` (finer voice/music separation).\n * `removeBackgroundNoise` additionally denoises the result.\n * `musicVolumeMode` sets the level of that preserved background (only relevant\n * when `preserveBackground` is on): `\"match\"` (default) keeps the original\n * level, `\"normalize\"` loudnorms it, `\"manual\"` uses `musicVolume`%.\n * `voiceFx` applies a reverb/echo to the COMBINED recast voices BEFORE the\n * background is mixed back in (effect sits on the voices, not the music bed).\n *\n * Cloud-only — costs credits and runs async; poll `jobs.get(jobId)` for the\n * result (`output_data.videoUrl` + `output_data.audioUrl` in video mode).\n */\n recast(input: VoiceChangerProInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro\", { body: input })\n }\n\n /**\n * Detect the speakers in a clip WITHOUT recasting yet\n * (`POST /v1/voice-changer-pro/analyze`) — the first step of the interactive\n * flow. Separates voice from music once and diarizes the vocals, returning the\n * speaker list so a user (or agent) can choose a voice per speaker before\n * committing to a paid recast. Poll `jobs.get(jobId)`: the completed job's\n * `output_data` carries the separated stem urls + the detected `speakers`\n * (each with `id`, time `segments`, `firstStartSec`, `wordCount`, `snippet`)\n * and the detected language — reshape it into a {@link VcpAnalysis} and pass it\n * as `recast({ ..., analysis })` to skip re-detection. With `suggestTitle`,\n * `output_data.suggestedTitle` also carries an LLM-proposed title.\n *\n * Cloud-only; costs credits and runs async.\n */\n analyze(input: VcpAnalyzeInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro/analyze\", { body: input })\n }\n\n /**\n * Render a final video from a mixed set of stems\n * (`POST /v1/voice-changer-pro/export`) — the last step of the interactive\n * flow. After `recast({ output: \"stems\" })` hands back the dry per-track stems\n * and the user has set levels / mutes / an effect in your editor, pass those\n * `tracks` (plus the source `videoUrl`) here to mix and remux into the finished\n * video. The video is stream-copied (never re-encoded), so the export is\n * bit-identical to your preview. At least one track must be un-muted (all-muted\n * is a 400); `voiceFx` is applied to the voice tracks at render time.\n *\n * Cloud-only; costs credits and runs async — poll `jobs.get(jobId)` for the\n * result (`output_data.videoUrl`).\n */\n exportMix(input: VcpExportInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro/export\", { body: input })\n }\n\n /**\n * Design a brand-new synthetic voice from a text description\n * (`POST /v1/voice-design`) — ElevenLabs text-to-voice. `text` (100–1000 chars)\n * is a preview line spoken in the designed voice; `voiceDescription` describes\n * the voice to create. Costs credits and runs async — poll `jobs.get(jobId)`\n * for the preview + the reusable voice id.\n */\n design(input: VoiceDesignInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-design\", { body: input })\n }\n\n /**\n * Generate speech in a voice described in natural language, without cloning\n * (`POST /v1/voice-remix`). `text` (1–5000 chars) is spoken in a voice matching\n * `voiceDescription`. Costs credits and runs async — poll `jobs.get(jobId)`.\n */\n remix(input: VoiceRemixInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-remix\", { body: input })\n }\n\n /**\n * Dub an audio clip into another language while preserving each speaker's voice\n * (`POST /v1/dubbing`). `targetLanguage` is an ISO code (e.g. `\"es\"`, `\"fr\"`);\n * `sourceLanguage` is auto-detected when omitted. Costs credits and runs async\n * — poll `jobs.get(jobId)`.\n */\n dub(input: DubbingInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/dubbing\", { body: input })\n }\n\n /**\n * Voice a multi-speaker script as ONE audio file\n * (`POST /v1/text-to-dialogue`) on ElevenLabs Dialogue v3 — the default — or\n * Dialogue v4 (`provider: \"elevenlabs-dialogue-v4\"`). Each `dialogue` line is\n * `{ text, voice }` in speaking order — `voice` is a premade voice name or an\n * ElevenLabs voice UUID (cloned/library voices work too; mixed casts are\n * fine). At most 5,000 characters total across lines on Dialogue v3 and\n * 10,000 on Dialogue v4 (≤2,000 recommended for best quality) and 10 unique\n * voices. Line text may\n * carry `[audio tags]` like `[laughs]`. Costs credits and runs async — poll\n * `jobs.get(jobId)` for `output_data.audioUrl` and — on every dialogue model,\n * since both return timings at no extra credits — `output_data.transcript`, a\n * `Transcript` (`@nodaro/shared`): `words[]` with `startMs` / `endMs` /\n * `speaker` and one `segments[]` entry per line, `speaker` being the line's\n * `voice` as you sent it. Pass `transcript.words` to\n * `addCaptions({ captions, autoTranscribe: false })` to caption the dialogue\n * without a second transcription.\n */\n textToDialogue(input: {\n dialogue: Array<{ text: string; voice: string }>\n /** Dialogue model: `\"elevenlabs-dialogue\"` (v3, the default) or `\"elevenlabs-dialogue-v4\"`. */\n provider?: DialogueProvider\n /** v3: exactly 0, 0.5 or 1 (0 = most variable, 1 = most stable). v4: any value 0–1. */\n stability?: number\n /** v4 only (v3 ignores it): how closely each line keeps its voice's character, 0–1. */\n similarityBoost?: number\n /** ISO 639-1 language hint (e.g. \"en\"); auto-detected when omitted. */\n languageCode?: string\n /** Deterministic sampling seed (integer 0–4294967295). Omit for random. */\n seed?: number\n /** Spell out numbers/dates/abbreviations: \"auto\" (default) | \"on\" | \"off\". */\n applyTextNormalization?: \"auto\" | \"on\" | \"off\"\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/text-to-dialogue\", { body: input })\n }\n\n /**\n * @deprecated Voice cloning is no longer offered on Nodaro (retired\n * 2026-09-15). The route answers `410 voice_cloning_retired` and this call\n * rejects with that error. See {@link VoicesResource.createClone}.\n */\n createCloneFromFile(input: {\n name: string\n file: Blob | Uint8Array | ArrayBuffer\n /** File name for the upload part (default `sample`). */\n filename?: string\n /** MIME type when `file` is a raw buffer (default `audio/mpeg`). */\n contentType?: string\n }): Promise<VoiceClone> {\n const form = new FormData()\n // Append the name field BEFORE the file so the route sees it in `fields`\n // (fastify-multipart's `req.file()` exposes fields parsed up to the file).\n form.append(\"name\", input.name)\n const blob =\n input.file instanceof Blob\n ? input.file\n : new Blob([input.file as BlobPart], { type: input.contentType ?? \"audio/mpeg\" })\n form.append(\"file\", blob, input.filename ?? \"sample\")\n return this.client.request<VoiceClone>(\"POST\", \"/v1/voice-clones\", { body: form })\n }\n}\n\n/**\n * One entry in {@link VoiceChangerProInput.orderedVoices}. Either a bare voice id\n * (premade name like `\"Rachel\"` or an ElevenLabs UUID for a custom clone), or\n * an object pinning per-voice speech-to-speech settings and the recast's\n * loudness behaviour for that speaker.\n */\nexport type VoiceChangerProVoice =\n | string\n | {\n /** Target voice — premade name (`\"Rachel\"`, `\"Aria\"`, …) or an ElevenLabs UUID for a custom clone. */\n voiceId: string\n /**\n * Which lane converts this speaker. `\"sts\"` (default) is the classic\n * speech-to-speech recast; `\"v3\"` and `\"v4\"` are RE-SPEAK — the\n * performance is REGENERATED from the transcript (supports `[audio\n * tags]`; original delivery is replaced, and lips won't match on video).\n * A Re-speak speaker needs transcript text: pass an `analysis` whose\n * `segments[].text` carries it (analyze now emits this), or omit\n * `analysis` and the engine re-speaks from its own transcription. For\n * `\"v3\"` (eleven_v3), `stability` accepts exactly 0 / 0.5 / 1, and\n * `similarityBoost` / `style` / `useSpeakerBoost` are ignored. For\n * `\"v4\"` (eleven_v4), `stability` takes any value 0–1 and\n * `similarityBoost` is honoured (`style` / `useSpeakerBoost` ignored);\n * each line is generated with its neighbouring lines as context for\n * smoother joins. Both Re-speak engines are priced the same, per started\n * 1,000 characters.\n */\n engine?: \"sts\" | \"v3\" | \"v4\"\n /** ElevenLabs stability (0–1). Higher = steadier, lower = more expressive. */\n stability?: number\n /** ElevenLabs similarity boost (0–1) — how closely the output hugs the target voice's timbre. */\n similarityBoost?: number\n /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency / stability. */\n style?: number\n /** ElevenLabs speaker boost — sharpens fidelity to the target speaker. */\n useSpeakerBoost?: boolean\n /**\n * Deterministic speech-to-speech seed (integer 0–4294967295) for\n * reproducible per-voice output — the same source + settings + seed\n * recast this speaker identically across runs. Omit for a random seed.\n */\n seed?: number\n /**\n * Loudness handling for this recast voice. `\"match\"` (default) matches the\n * original speaker's loudness; `\"normalize\"` applies loudness\n * normalization; `\"manual\"` uses `volume` as a percentage.\n */\n volumeMode?: \"match\" | \"normalize\" | \"manual\"\n /** Manual output volume as a percentage (0–200). Consulted only when `volumeMode === \"manual\"`. */\n volume?: number\n }\n\n/** Input for {@link VoicesResource.recast}. */\nexport interface VoiceChangerProInput {\n /** URL of an audio file to recast (audio → audio). Exactly one of `audioUrl` / `videoUrl` is required. */\n audioUrl?: string\n /** URL of a video file to recast (the audio track is recast and remuxed). Exactly one of `audioUrl` / `videoUrl` is required. */\n videoUrl?: string\n /**\n * Voices in speaker-detection order. Speaker N is mapped to `orderedVoices[N]`;\n * speakers beyond the array keep their original voice. Each entry is a bare\n * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings —\n * OR `null`, meaning keep this speaker's original voice (requires a platform\n * running cloud-plugins with keep-slot support). At least one entry must be\n * non-null.\n */\n orderedVoices: Array<VoiceChangerProVoice | null>\n /** Model to use for speech-to-speech. Defaults to the server-configured default when omitted. */\n model?: string\n /**\n * Mix the separated music / SFX stem back under the recast voices. Default\n * `true`. The voice is ALWAYS split out before recasting regardless of this\n * flag — `false` simply drops the music for a clean voice-only result.\n */\n preserveBackground?: boolean\n /**\n * Quality of the voice/music separation. `\"fast\"` (default, quicker —\n * preserves more of the voice) or `\"best\"` (finer voice/music separation).\n */\n separationQuality?: \"fast\" | \"best\"\n /** Strip background noise for a clean voice-only result. */\n removeBackgroundNoise?: boolean\n /**\n * Level of the preserved background music / SFX stem in the final mix. Only\n * relevant when `preserveBackground` is on (otherwise there is no background to\n * level). `\"match\"` (default) leaves the separated instrumental at its original\n * level; `\"normalize\"` applies loudness normalization; `\"manual\"` sets its\n * level to `musicVolume`%.\n */\n musicVolumeMode?: \"match\" | \"normalize\" | \"manual\"\n /** Background music level as a percentage (0–200). Consulted only when `musicVolumeMode === \"manual\"`. */\n musicVolume?: number\n /**\n * Node-level reverb/echo applied to the COMBINED recast voices **before** the\n * background is mixed back in (so the effect sits on the voices only, not the\n * music/SFX bed). Reverb presets (`room`, `hall`, `church`, …) use\n * `wetDryMix`; the `echo` / `custom` presets use `delayMs` + `decay`. Omit for\n * no effect.\n */\n voiceFx?: {\n /** Effect preset — reverb space, `telephone`, `megaphone`, `echo`, or `custom`. */\n preset: AudioFxPreset\n /** Reverb wet/dry mix as a percentage (0–100). Higher = wetter (more reverb). */\n wetDryMix?: number\n /** Echo delay in milliseconds (20–2000). Used by the `echo` / `custom` presets. */\n delayMs?: number\n /** Echo decay / feedback (0–1). Higher = more repeats. Used by the `echo` / `custom` presets. */\n decay?: number\n }\n /**\n * Output mode. `\"video\"` (default) mixes the recast voices with the preserved\n * background and returns a finished merged video. `\"stems\"` returns the dry,\n * unleveled per-track stems instead (rendering nothing) so you can drive an\n * INTERACTIVE mix — adjust levels/mutes/effect in your own UI, then render the\n * final video with {@link VoicesResource.exportMix}. This is how an app builds\n * a full editor around VCP rather than a one-shot recast.\n */\n output?: \"video\" | \"stems\"\n /**\n * A prior {@link VoicesResource.analyze} result. Pass it to SKIP re-detection:\n * the recast reuses the already-separated stems and speaker segments instead of\n * running separation + diarization again. This is the fast-path for the\n * detect → pick voices → recast interactive flow (analyze once, recast N times\n * as the user tweaks voice assignments). Omit to auto-detect from the source.\n */\n analysis?: VcpAnalysis\n}\n\n/** One detected speaker in a {@link VcpAnalysis} (from `analyze`). */\nexport interface VcpAnalysisSpeaker {\n /** Stable speaker id (first-appearance order). */\n id: string\n /** The speaker's spoken time ranges (seconds). `text` is what was said in\n * the range — the paid input for a Re-speak speaker (`engine: \"v3\"` or\n * `\"v4\"`; editable before conversion); the STS lane ignores it. */\n segments: Array<{ start: number; end: number; text?: string }>\n /** When the speaker first speaks (seconds). */\n firstStartSec?: number\n /** Rough word count across the clip — a proxy for how much this speaker says. */\n wordCount?: number\n /** The first few transcribed words, to help a user tell speakers apart. */\n snippet?: string\n}\n\n/**\n * The result of {@link VoicesResource.analyze}, reshaped to pass back into\n * {@link VoiceChangerProInput.analysis}. Read a completed analyze job's\n * `output_data` into this shape (it carries the separated stem urls + the\n * detected speakers) and thread it into `recast` to skip re-detection.\n */\nexport interface VcpAnalysis {\n /** URL of the isolated vocal stem. */\n vocalsUrl: string\n /** URL of the separated music/SFX stem (absent when the source had none). */\n backgroundUrl?: string\n /** The detected speakers, in first-appearance order. */\n speakers: VcpAnalysisSpeaker[]\n /** Scribe's detected language code, round-tripped so the recast auto-selects the STS model. */\n languageCode?: string\n /** Confidence (0–1) of {@link VcpAnalysis.languageCode}. */\n languageProbability?: number\n}\n\n/** Input for {@link VoicesResource.analyze}. */\nexport interface VcpAnalyzeInput {\n /** URL of an audio file to analyze. Exactly one of `audioUrl` / `videoUrl` is required. */\n audioUrl?: string\n /** URL of a video file to analyze (its audio track is used). Exactly one of `audioUrl` / `videoUrl` is required. */\n videoUrl?: string\n /** Quality of the voice/music separation run before diarization: `\"fast\"` (default) or `\"best\"`. */\n separationQuality?: \"fast\" | \"best\"\n /** Also suggest a conversion title from the transcript (returned on the job's `output_data.suggestedTitle`). */\n suggestTitle?: boolean\n}\n\n/** One track in a {@link VcpExportInput} mix. */\nexport interface VcpExportTrack {\n /** URL of the stem for this lane (a recast voice stem or the background stem). */\n url: string\n /** Fader position as a percentage: 0 = silent, 100 = unity, 200 = +6dB. */\n gain: number\n /** Whether this lane is muted in the mix. */\n muted: boolean\n /**\n * Which bucket the track is in, and so whether `voiceFx` lands on it. Defaults\n * to `\"voice\"`. Set `\"background\"` for the music/SFX lane (the effect never\n * touches it).\n */\n kind?: \"voice\" | \"background\"\n}\n\n/** Input for {@link VoicesResource.exportMix}. */\nexport interface VcpExportInput {\n /** The source video to remux the mixed audio onto (stream-copied — never re-encoded). */\n videoUrl: string\n /** The mix: one entry per lane. At least one must be un-muted (all-muted is a 400). Max 16 tracks. */\n tracks: VcpExportTrack[]\n /**\n * A reverb/echo applied to the VOICE tracks only (not `\"background\"` lanes)\n * at render time — so iterating the effect in your editor is free until you\n * export. Same shape as {@link VoiceChangerProInput.voiceFx}.\n */\n voiceFx?: VoiceChangerProInput[\"voiceFx\"]\n}\n\n/** Input for {@link VoicesResource.design}. */\nexport interface VoiceDesignInput {\n /** A preview line (100–1000 chars) spoken in the designed voice. */\n text: string\n /** Natural-language description of the voice to create. */\n voiceDescription: string\n /** Voice-design model id. Defaults to the server-configured default. */\n model?: string\n /** Output loudness (-1..1). */\n loudness?: number\n /** How strongly the description steers the design (0–100). */\n guidanceScale?: number\n /** Deterministic seed for a reproducible design. */\n seed?: number\n /** Design quality knob (provider-specific). */\n quality?: number\n /** Enhance the generated voice. */\n shouldEnhance?: boolean\n /** Optional extra prompt context (≤8000 chars). */\n userPrompt?: string\n}\n\n/** Input for {@link VoicesResource.remix}. */\nexport interface VoiceRemixInput {\n /** The text (1–5000 chars) to speak in the described voice. */\n text: string\n /** Natural-language description of the voice to speak in. */\n voiceDescription: string\n /** Optional extra prompt context (≤8000 chars). */\n userPrompt?: string\n}\n\n/** Input for {@link VoicesResource.dub}. Provide exactly ONE source:\n * `audioUrl` (dubbed audio out), `videoUrl` (the dubbed VIDEO out, plus the\n * dubbed audio track as `output_data.audioUrl`), or `sourceUrl` (a public\n * YouTube/TikTok/direct link ElevenLabs fetches itself). Priced per minute\n * of the dubbed span; the span is capped at 30 minutes — use\n * `startTime`/`endTime` to dub part of a longer source. */\nexport interface DubbingInput {\n /** URL of the audio to dub (audio mode). */\n audioUrl?: string\n /** URL of the video to dub — the result is the dubbed VIDEO (+ audio track). */\n videoUrl?: string\n /** Public page/media URL (YouTube, TikTok, or a direct link) — ElevenLabs fetches it directly. */\n sourceUrl?: string\n /** Target language ISO code (2–10 chars), e.g. `\"es\"`, `\"pt-BR\"`. */\n targetLanguage: string\n /** Source language ISO code; auto-detected when omitted. */\n sourceLanguage?: string\n /** Expected number of speakers (1–20; 0 = auto-detect) — improves separation when known. */\n numSpeakers?: number\n /** Keep the original voices instead of cloning them into the target language. */\n disableVoiceCloning?: boolean\n /** Drop the background/music bed from the dubbed output. */\n dropBackgroundAudio?: boolean\n /** Dub only this window of the source (seconds). */\n startTime?: number\n endTime?: number\n /** Keep the source resolution on video dubs (slower render). */\n highestResolution?: boolean\n useProfanityFilter?: boolean\n /** Experimental: steer dubbed voices toward an accent. */\n targetAccent?: string\n /** Apply ElevenLabs' own watermark to video dubs. */\n watermark?: boolean\n}\n"]}
1
+ {"version":3,"sources":["../src/resources/voices.ts"],"names":[],"mappings":";AAoBO,IAAM,iBAAN,MAAqB;AAAA,EAC1B,YAAoB,MAAA,EAA0B;AAA1B,IAAA,IAAA,CAAA,MAAA,GAAA,MAAA;AAAA,EAA2B;AAAA,EAA3B,MAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMpB,MAAM,IAAA,GAAyB;AAC7B,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAA6B,OAAO,YAAY,CAAA;AAC9E,IAAA,OAAO,GAAA,CAAI,MAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,aAAA,CAAc,MAAA,GAA6B,EAAC,EAAkC;AAC5E,IAAA,MAAM,QAA+D,EAAC;AACtE,IAAA,KAAA,MAAW,CAAC,CAAA,EAAG,CAAC,KAAK,MAAA,CAAO,OAAA,CAAQ,MAAM,CAAA,EAAG;AAC3C,MAAA,IAAI,CAAA,KAAM,UAAa,CAAA,KAAM,IAAA,IAAQ,MAAM,EAAA,EAAI,KAAA,CAAM,CAAC,CAAA,GAAI,CAAA;AAAA,IAC5D;AACA,IAAA,OAAO,KAAK,MAAA,CAAO,OAAA,CAAQ,OAAO,oBAAA,EAAsB,EAAE,OAAO,CAAA;AAAA,EACnE;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,UAAA,GAAoC;AACxC,IAAA,MAAM,MAAM,MAAM,IAAA,CAAK,MAAA,CAAO,OAAA,CAAuC,OAAO,kBAAkB,CAAA;AAC9F,IAAA,OAAO,GAAA,CAAI,WAAA;AAAA,EACb;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,YAAY,KAAA,EAAgE;AAC1E,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAoB,MAAA,EAAQ,6BAA6B,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC7F;AAAA;AAAA,EAGA,MAAM,YAAY,EAAA,EAA2B;AAC3C,IAAA,MAAM,IAAA,CAAK,OAAO,OAAA,CAAc,QAAA,EAAU,oBAAoB,kBAAA,CAAmB,EAAE,CAAC,CAAA,CAAE,CAAA;AAAA,EACxF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAaA,OAAO,KAAA,EAewB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,qBAAqB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC5F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoCA,OAAO,KAAA,EAAyD;AAC9D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,yBAAyB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAChG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBA,QAAQ,KAAA,EAAoD;AAC1D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,iCAAiC,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACxG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAeA,UAAU,KAAA,EAAmD;AAC3D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,gCAAgC,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACvG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAO,KAAA,EAAqD;AAC1D,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,oBAAoB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC3F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,KAAA,EAAoD;AACxD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,mBAAmB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC1F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,IAAI,KAAA,EAAiD;AACnD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,eAAe,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EACtF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,eAAe,KAAA,EAcgB;AAC7B,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAA2B,MAAA,EAAQ,wBAAwB,EAAE,IAAA,EAAM,OAAO,CAAA;AAAA,EAC/F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,oBAAoB,KAAA,EAOI;AACtB,IAAA,MAAM,IAAA,GAAO,IAAI,QAAA,EAAS;AAG1B,IAAA,IAAA,CAAK,MAAA,CAAO,MAAA,EAAQ,KAAA,CAAM,IAAI,CAAA;AAC9B,IAAA,MAAM,OACJ,KAAA,CAAM,IAAA,YAAgB,IAAA,GAClB,KAAA,CAAM,OACN,IAAI,IAAA,CAAK,CAAC,KAAA,CAAM,IAAgB,CAAA,EAAG,EAAE,MAAM,KAAA,CAAM,WAAA,IAAe,cAAc,CAAA;AACpF,IAAA,IAAA,CAAK,MAAA,CAAO,MAAA,EAAQ,IAAA,EAAM,KAAA,CAAM,YAAY,QAAQ,CAAA;AACpD,IAAA,OAAO,IAAA,CAAK,OAAO,OAAA,CAAoB,MAAA,EAAQ,oBAAoB,EAAE,IAAA,EAAM,MAAM,CAAA;AAAA,EACnF;AACF","file":"chunk-CRIQLLR2.js","sourcesContent":["import type { Voice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from \"@nodaro/shared/pipeline\"\nimport type { AudioFxPreset } from \"@nodaro/shared/media\"\nimport type { DialogueProvider } from \"@nodaro/shared/models\"\nexport type { Voice, SharedVoice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from \"@nodaro/shared/pipeline\"\nexport type { AudioFxPreset } from \"@nodaro/shared/media\"\nimport type { NodaroClientCore } from \"../core.js\"\n\n/**\n * Re-export the shared voice types so SDK consumers don't have to add\n * `@nodaro/shared` as a second dependency just to type a `Voice` row, a\n * `VoiceClone`, or a `searchLibrary` call. Single source of truth lives in\n * `@nodaro/shared`.\n */\n/** Audio-FX preset union (reverb spaces / telephone / megaphone / echo / custom) — used by {@link VoiceChangerProInput.voiceFx}. */\n\n/**\n * Read access to ElevenLabs voices: the premade catalog plus the shared\n * community Voice Library (both public GETs, no body), and the signed-in\n * user's own pre-retirement voice clones (list / delete).\n */\nexport class VoicesResource {\n constructor(private client: NodaroClientCore) {}\n\n /**\n * List the premade ElevenLabs voices (`GET /v1/voices`). Falls back to a\n * curated set server-side when no ElevenLabs API key is configured.\n */\n async list(): Promise<Voice[]> {\n const res = await this.client.request<{ voices: Voice[] }>(\"GET\", \"/v1/voices\")\n return res.voices\n }\n\n /**\n * Search the shared/community Voice Library (`GET /v1/voices/library`). All\n * params are optional and forwarded as a querystring; `undefined` / `null` /\n * empty-string values are omitted so the server defaults apply. `hasMore`\n * drives \"load more\" pagination.\n */\n searchLibrary(params: VoiceLibraryParams = {}): Promise<VoiceLibraryResponse> {\n const query: Record<string, string | number | boolean | undefined> = {}\n for (const [k, v] of Object.entries(params)) {\n if (v !== undefined && v !== null && v !== \"\") query[k] = v as string | number | boolean\n }\n return this.client.request(\"GET\", \"/v1/voices/library\", { query })\n }\n\n /**\n * List the signed-in user's voice clones (`GET /v1/voice-clones`). The route\n * wraps the rows in `{ voiceClones }`; we unwrap to the bare array.\n */\n async listClones(): Promise<VoiceClone[]> {\n const res = await this.client.request<{ voiceClones: VoiceClone[] }>(\"GET\", \"/v1/voice-clones\")\n return res.voiceClones\n }\n\n /**\n * @deprecated Voice cloning is no longer offered on Nodaro (retired\n * 2026-09-15). The route answers `410 voice_cloning_retired` and this call\n * rejects with that error. Clones created before the retirement are still\n * listed by {@link VoicesResource.listClones} and still work as voice ids.\n * For a new custom voice use {@link VoicesResource.design}.\n */\n createClone(input: { name: string; audioUrl: string }): Promise<VoiceClone> {\n return this.client.request<VoiceClone>(\"POST\", \"/v1/voice-clones/from-url\", { body: input })\n }\n\n /** Delete one of the user's voice clones (`DELETE /v1/voice-clones/:id`). */\n async deleteClone(id: string): Promise<void> {\n await this.client.request<void>(\"DELETE\", `/v1/voice-clones/${encodeURIComponent(id)}`)\n }\n\n /**\n * Replace the voice in a recording — or in a whole talking video — with a\n * different voice (`POST /v1/voice-changer`). Pass `audioUrl` to revoice\n * audio→audio, or `videoUrl` to revoice an entire clip (the server demuxes\n * the audio, runs speech-to-speech, and remuxes onto the original video,\n * returning the video plus the new audio track). Exactly one of `audioUrl` /\n * `videoUrl` is required; when both are sent, video wins. `removeBackgroundNoise`\n * off keeps the music/SFX bed under the new voice; on yields a clean voice-only\n * result. Costs credits and runs async — poll `jobs.get(jobId)` for the result\n * (`output_data.videoUrl` + `output_data.audioUrl` in video mode).\n */\n change(input: {\n voiceId: string\n audioUrl?: string\n videoUrl?: string\n /** Speech-to-speech model id. Defaults to the server-configured default when omitted. */\n model?: string\n stability?: number\n similarityBoost?: number\n /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency/stability. */\n style?: number\n /** ElevenLabs speaker boost — sharpens fidelity to the target speaker (small latency cost). */\n useSpeakerBoost?: boolean\n /** Deterministic speech-to-speech seed (integer 0–4294967295) for reproducible output. Omit for random. */\n seed?: number\n removeBackgroundNoise?: boolean\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer\", { body: input })\n }\n\n /**\n * Recast each detected speaker in a multi-speaker recording to a different\n * voice (`POST /v1/voice-changer-pro`). `orderedVoices` maps speaker positions to\n * voices in detection order — speaker 0 → `orderedVoices[0]`, speaker 1 →\n * `orderedVoices[1]`, etc. Speakers beyond the end of `orderedVoices` keep\n * their original voice. Each entry is EITHER a bare voice id (premade name or\n * ElevenLabs UUID) OR a {@link VoiceChangerProVoice} object carrying per-voice\n * ElevenLabs speech-to-speech settings (stability / similarityBoost / style /\n * useSpeakerBoost / `seed`) plus a loudness `volumeMode` (and a manual\n * `volume`) — OR `null`, meaning keep this speaker's original voice (requires\n * a platform running cloud-plugins with keep-slot support). At least one\n * entry must be non-null. A per-voice `seed` makes that speaker's recast\n * reproducible.\n *\n * Pass `audioUrl` for audio-only recast or `videoUrl` to recast the audio\n * track of a video clip (the server demuxes, recasts, and remuxes).\n *\n * Voice and music are ALWAYS separated first — before recasting, the source\n * is split into an isolated vocal stem and a music/SFX stem.\n * `preserveBackground` (default `true`) only controls whether that\n * music/instrumental stem is mixed back under the new voices; set it `false`\n * for a clean voice-only result. `separationQuality` selects the quality of\n * the voice/music separation: `\"fast\"` (default, quicker — preserves more of\n * the voice) or `\"best\"` (finer voice/music separation).\n * `removeBackgroundNoise` additionally denoises the result.\n * `musicVolumeMode` sets the level of that preserved background (only relevant\n * when `preserveBackground` is on): `\"match\"` (default) keeps the original\n * level, `\"normalize\"` loudnorms it, `\"manual\"` uses `musicVolume`%.\n * `voiceFx` applies a reverb/echo to the COMBINED recast voices BEFORE the\n * background is mixed back in (effect sits on the voices, not the music bed).\n *\n * Cloud-only — costs credits and runs async; poll `jobs.get(jobId)` for the\n * result (`output_data.videoUrl` + `output_data.audioUrl` in video mode).\n */\n recast(input: VoiceChangerProInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro\", { body: input })\n }\n\n /**\n * Detect the speakers in a clip WITHOUT recasting yet\n * (`POST /v1/voice-changer-pro/analyze`) — the first step of the interactive\n * flow. Separates voice from music once and diarizes the vocals, returning the\n * speaker list so a user (or agent) can choose a voice per speaker before\n * committing to a paid recast. Poll `jobs.get(jobId)`: the completed job's\n * `output_data` carries the separated stem urls + the detected `speakers`\n * (each with `id`, time `segments`, `firstStartSec`, `wordCount`, `snippet`)\n * and the detected language — reshape it into a {@link VcpAnalysis} and pass it\n * as `recast({ ..., analysis })` to skip re-detection. With `suggestTitle`,\n * `output_data.suggestedTitle` also carries an LLM-proposed title.\n *\n * Cloud-only; costs credits and runs async.\n */\n analyze(input: VcpAnalyzeInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro/analyze\", { body: input })\n }\n\n /**\n * Render a final video from a mixed set of stems\n * (`POST /v1/voice-changer-pro/export`) — the last step of the interactive\n * flow. After `recast({ output: \"stems\" })` hands back the dry per-track stems\n * and the user has set levels / mutes / an effect in your editor, pass those\n * `tracks` (plus the source `videoUrl`) here to mix and remux into the finished\n * video. The video is stream-copied (never re-encoded), so the export is\n * bit-identical to your preview. At least one track must be un-muted (all-muted\n * is a 400); `voiceFx` is applied to the voice tracks at render time.\n *\n * Cloud-only; costs credits and runs async — poll `jobs.get(jobId)` for the\n * result (`output_data.videoUrl`).\n */\n exportMix(input: VcpExportInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-changer-pro/export\", { body: input })\n }\n\n /**\n * Design a brand-new synthetic voice from a text description\n * (`POST /v1/voice-design`) — ElevenLabs text-to-voice. `text` (100–1000 chars)\n * is a preview line spoken in the designed voice; `voiceDescription` describes\n * the voice to create. Costs credits and runs async — poll `jobs.get(jobId)`\n * for the preview + the reusable voice id.\n */\n design(input: VoiceDesignInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-design\", { body: input })\n }\n\n /**\n * Generate speech in a voice described in natural language, without cloning\n * (`POST /v1/voice-remix`). `text` (1–5000 chars) is spoken in a voice matching\n * `voiceDescription`. Costs credits and runs async — poll `jobs.get(jobId)`.\n */\n remix(input: VoiceRemixInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/voice-remix\", { body: input })\n }\n\n /**\n * Dub an audio clip into another language while preserving each speaker's voice\n * (`POST /v1/dubbing`). `targetLanguage` is an ISO code (e.g. `\"es\"`, `\"fr\"`);\n * `sourceLanguage` is auto-detected when omitted. Costs credits and runs async\n * — poll `jobs.get(jobId)`.\n */\n dub(input: DubbingInput): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/dubbing\", { body: input })\n }\n\n /**\n * Voice a multi-speaker script as ONE audio file\n * (`POST /v1/text-to-dialogue`) on ElevenLabs Dialogue v3 — the default — or\n * Dialogue v4 (`provider: \"elevenlabs-dialogue-v4\"`). Each `dialogue` line is\n * `{ text, voice }` in speaking order — `voice` is a premade voice name or an\n * ElevenLabs voice UUID (cloned/library voices work too; mixed casts are\n * fine). At most 5,000 characters total across lines on Dialogue v3 and\n * 10,000 on Dialogue v4 (≤2,000 recommended for best quality) and 10 unique\n * voices. Line text may\n * carry `[audio tags]` like `[laughs]`. Costs credits and runs async — poll\n * `jobs.get(jobId)` for `output_data.audioUrl` and — on every dialogue model,\n * since both return timings at no extra credits — `output_data.transcript`, a\n * `Transcript` (`@nodaro/shared`): `words[]` with `startMs` / `endMs` /\n * `speaker` and one `segments[]` entry per line, `speaker` being the line's\n * `voice` as you sent it. Pass `transcript.words` to\n * `addCaptions({ captions, autoTranscribe: false })` to caption the dialogue\n * without a second transcription.\n */\n textToDialogue(input: {\n dialogue: Array<{ text: string; voice: string }>\n /** Dialogue model: `\"elevenlabs-dialogue\"` (v3, the default) or `\"elevenlabs-dialogue-v4\"`. */\n provider?: DialogueProvider\n /** v3: exactly 0, 0.5 or 1 (0 = most variable, 1 = most stable). v4: any value 0–1. */\n stability?: number\n /** v4 only (v3 ignores it): how closely each line keeps its voice's character, 0–1. */\n similarityBoost?: number\n /** ISO 639-1 language hint (e.g. \"en\"); auto-detected when omitted. */\n languageCode?: string\n /** Deterministic sampling seed (integer 0–4294967295). Omit for random. */\n seed?: number\n /** Spell out numbers/dates/abbreviations: \"auto\" (default) | \"on\" | \"off\". */\n applyTextNormalization?: \"auto\" | \"on\" | \"off\"\n }): Promise<{ jobId: string }> {\n return this.client.request<{ jobId: string }>(\"POST\", \"/v1/text-to-dialogue\", { body: input })\n }\n\n /**\n * @deprecated Voice cloning is no longer offered on Nodaro (retired\n * 2026-09-15). The route answers `410 voice_cloning_retired` and this call\n * rejects with that error. See {@link VoicesResource.createClone}.\n */\n createCloneFromFile(input: {\n name: string\n file: Blob | Uint8Array | ArrayBuffer\n /** File name for the upload part (default `sample`). */\n filename?: string\n /** MIME type when `file` is a raw buffer (default `audio/mpeg`). */\n contentType?: string\n }): Promise<VoiceClone> {\n const form = new FormData()\n // Append the name field BEFORE the file so the route sees it in `fields`\n // (fastify-multipart's `req.file()` exposes fields parsed up to the file).\n form.append(\"name\", input.name)\n const blob =\n input.file instanceof Blob\n ? input.file\n : new Blob([input.file as BlobPart], { type: input.contentType ?? \"audio/mpeg\" })\n form.append(\"file\", blob, input.filename ?? \"sample\")\n return this.client.request<VoiceClone>(\"POST\", \"/v1/voice-clones\", { body: form })\n }\n}\n\n/**\n * One entry in {@link VoiceChangerProInput.orderedVoices}. Either a bare voice id\n * (premade name like `\"Rachel\"` or an ElevenLabs UUID for a custom clone), or\n * an object pinning per-voice speech-to-speech settings and the recast's\n * loudness behaviour for that speaker.\n */\nexport type VoiceChangerProVoice =\n | string\n | {\n /** Target voice — premade name (`\"Rachel\"`, `\"Aria\"`, …) or an ElevenLabs UUID for a custom clone. */\n voiceId: string\n /**\n * Which lane converts this speaker. `\"sts\"` (default) is the classic\n * speech-to-speech recast; `\"v3\"` and `\"v4\"` are RE-SPEAK — the\n * performance is REGENERATED from the transcript (supports `[audio\n * tags]`; original delivery is replaced, and lips won't match on video).\n * A Re-speak speaker needs transcript text: pass an `analysis` whose\n * `segments[].text` carries it (analyze now emits this), or omit\n * `analysis` and the engine re-speaks from its own transcription. For\n * `\"v3\"` (eleven_v3), `stability` accepts exactly 0 / 0.5 / 1, and\n * `similarityBoost` / `style` / `useSpeakerBoost` are ignored. For\n * `\"v4\"` (eleven_v4), `stability` takes any value 0–1 and\n * `similarityBoost` is honoured (`style` / `useSpeakerBoost` ignored);\n * each line is generated with its neighbouring lines as context for\n * smoother joins. Both Re-speak engines are priced the same, per started\n * 1,000 characters.\n */\n engine?: \"sts\" | \"v3\" | \"v4\"\n /** ElevenLabs stability (0–1). Higher = steadier, lower = more expressive. */\n stability?: number\n /** ElevenLabs similarity boost (0–1) — how closely the output hugs the target voice's timbre. */\n similarityBoost?: number\n /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency / stability. */\n style?: number\n /** ElevenLabs speaker boost — sharpens fidelity to the target speaker. */\n useSpeakerBoost?: boolean\n /**\n * Deterministic speech-to-speech seed (integer 0–4294967295) for\n * reproducible per-voice output — the same source + settings + seed\n * recast this speaker identically across runs. Omit for a random seed.\n */\n seed?: number\n /**\n * Loudness handling for this recast voice. `\"match\"` (default) matches the\n * original speaker's loudness; `\"normalize\"` applies loudness\n * normalization; `\"manual\"` uses `volume` as a percentage.\n */\n volumeMode?: \"match\" | \"normalize\" | \"manual\"\n /** Manual output volume as a percentage (0–200). Consulted only when `volumeMode === \"manual\"`. */\n volume?: number\n }\n\n/** Input for {@link VoicesResource.recast}. */\nexport interface VoiceChangerProInput {\n /** URL of an audio file to recast (audio → audio). Exactly one of `audioUrl` / `videoUrl` is required. */\n audioUrl?: string\n /** URL of a video file to recast (the audio track is recast and remuxed). Exactly one of `audioUrl` / `videoUrl` is required. */\n videoUrl?: string\n /**\n * Voices in speaker-detection order. Speaker N is mapped to `orderedVoices[N]`;\n * speakers beyond the array keep their original voice. Each entry is a bare\n * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings —\n * OR `null`, meaning keep this speaker's original voice (requires a platform\n * running cloud-plugins with keep-slot support). At least one entry must be\n * non-null.\n */\n orderedVoices: Array<VoiceChangerProVoice | null>\n /** Model to use for speech-to-speech. Defaults to the server-configured default when omitted. */\n model?: string\n /**\n * Mix the separated music / SFX stem back under the recast voices. Default\n * `true`. The voice is ALWAYS split out before recasting regardless of this\n * flag — `false` simply drops the music for a clean voice-only result.\n */\n preserveBackground?: boolean\n /**\n * Quality of the voice/music separation. `\"fast\"` (default, quicker —\n * preserves more of the voice) or `\"best\"` (finer voice/music separation).\n */\n separationQuality?: \"fast\" | \"best\"\n /** Strip background noise for a clean voice-only result. */\n removeBackgroundNoise?: boolean\n /**\n * Level of the preserved background music / SFX stem in the final mix. Only\n * relevant when `preserveBackground` is on (otherwise there is no background to\n * level). `\"match\"` (default) leaves the separated instrumental at its original\n * level; `\"normalize\"` applies loudness normalization; `\"manual\"` sets its\n * level to `musicVolume`%.\n */\n musicVolumeMode?: \"match\" | \"normalize\" | \"manual\"\n /** Background music level as a percentage (0–200). Consulted only when `musicVolumeMode === \"manual\"`. */\n musicVolume?: number\n /**\n * Node-level reverb/echo applied to the COMBINED recast voices **before** the\n * background is mixed back in (so the effect sits on the voices only, not the\n * music/SFX bed). Reverb presets (`room`, `hall`, `church`, …) use\n * `wetDryMix`; the `echo` / `custom` presets use `delayMs` + `decay`. Omit for\n * no effect.\n */\n voiceFx?: {\n /** Effect preset — reverb space, `telephone`, `megaphone`, `echo`, or `custom`. */\n preset: AudioFxPreset\n /** Reverb wet/dry mix as a percentage (0–100). Higher = wetter (more reverb). */\n wetDryMix?: number\n /** Echo delay in milliseconds (20–2000). Used by the `echo` / `custom` presets. */\n delayMs?: number\n /** Echo decay / feedback (0–1). Higher = more repeats. Used by the `echo` / `custom` presets. */\n decay?: number\n }\n /**\n * Output mode. `\"video\"` (default) mixes the recast voices with the preserved\n * background and returns a finished merged video. `\"stems\"` returns the dry,\n * unleveled per-track stems instead (rendering nothing) so you can drive an\n * INTERACTIVE mix — adjust levels/mutes/effect in your own UI, then render the\n * final video with {@link VoicesResource.exportMix}. This is how an app builds\n * a full editor around VCP rather than a one-shot recast.\n */\n output?: \"video\" | \"stems\"\n /**\n * A prior {@link VoicesResource.analyze} result. Pass it to SKIP re-detection:\n * the recast reuses the already-separated stems and speaker segments instead of\n * running separation + diarization again. This is the fast-path for the\n * detect → pick voices → recast interactive flow (analyze once, recast N times\n * as the user tweaks voice assignments). Omit to auto-detect from the source.\n */\n analysis?: VcpAnalysis\n}\n\n/** One detected speaker in a {@link VcpAnalysis} (from `analyze`). */\nexport interface VcpAnalysisSpeaker {\n /** Stable speaker id (first-appearance order). */\n id: string\n /** The speaker's spoken time ranges (seconds). `text` is what was said in\n * the range — the paid input for a Re-speak speaker (`engine: \"v3\"` or\n * `\"v4\"`; editable before conversion); the STS lane ignores it. */\n segments: Array<{ start: number; end: number; text?: string }>\n /** When the speaker first speaks (seconds). */\n firstStartSec?: number\n /** Rough word count across the clip — a proxy for how much this speaker says. */\n wordCount?: number\n /** The first few transcribed words, to help a user tell speakers apart. */\n snippet?: string\n}\n\n/**\n * The result of {@link VoicesResource.analyze}, reshaped to pass back into\n * {@link VoiceChangerProInput.analysis}. Read a completed analyze job's\n * `output_data` into this shape (it carries the separated stem urls + the\n * detected speakers) and thread it into `recast` to skip re-detection.\n */\nexport interface VcpAnalysis {\n /** URL of the isolated vocal stem. */\n vocalsUrl: string\n /** URL of the separated music/SFX stem (absent when the source had none). */\n backgroundUrl?: string\n /** The detected speakers, in first-appearance order. */\n speakers: VcpAnalysisSpeaker[]\n /** Scribe's detected language code, round-tripped so the recast auto-selects the STS model. */\n languageCode?: string\n /** Confidence (0–1) of {@link VcpAnalysis.languageCode}. */\n languageProbability?: number\n}\n\n/** Input for {@link VoicesResource.analyze}. */\nexport interface VcpAnalyzeInput {\n /** URL of an audio file to analyze. Exactly one of `audioUrl` / `videoUrl` is required. */\n audioUrl?: string\n /** URL of a video file to analyze (its audio track is used). Exactly one of `audioUrl` / `videoUrl` is required. */\n videoUrl?: string\n /** Quality of the voice/music separation run before diarization: `\"fast\"` (default) or `\"best\"`. */\n separationQuality?: \"fast\" | \"best\"\n /** Also suggest a conversion title from the transcript (returned on the job's `output_data.suggestedTitle`). */\n suggestTitle?: boolean\n}\n\n/** One track in a {@link VcpExportInput} mix. */\nexport interface VcpExportTrack {\n /** URL of the stem for this lane (a recast voice stem or the background stem). */\n url: string\n /** Fader position as a percentage: 0 = silent, 100 = unity, 200 = +6dB. */\n gain: number\n /** Whether this lane is muted in the mix. */\n muted: boolean\n /**\n * Which bucket the track is in, and so whether `voiceFx` lands on it. Defaults\n * to `\"voice\"`. Set `\"background\"` for the music/SFX lane (the effect never\n * touches it).\n */\n kind?: \"voice\" | \"background\"\n}\n\n/** Input for {@link VoicesResource.exportMix}. */\nexport interface VcpExportInput {\n /** The source video to remux the mixed audio onto (stream-copied — never re-encoded). */\n videoUrl: string\n /** The mix: one entry per lane. At least one must be un-muted (all-muted is a 400). Max 16 tracks. */\n tracks: VcpExportTrack[]\n /**\n * A reverb/echo applied to the VOICE tracks only (not `\"background\"` lanes)\n * at render time — so iterating the effect in your editor is free until you\n * export. Same shape as {@link VoiceChangerProInput.voiceFx}.\n */\n voiceFx?: VoiceChangerProInput[\"voiceFx\"]\n}\n\n/** Input for {@link VoicesResource.design}. */\nexport interface VoiceDesignInput {\n /** A preview line (100–1000 chars) spoken in the designed voice. */\n text: string\n /** Natural-language description of the voice to create. */\n voiceDescription: string\n /** Voice-design model id. Defaults to the server-configured default. */\n model?: string\n /** Output loudness (-1..1). */\n loudness?: number\n /** How strongly the description steers the design (0–100). */\n guidanceScale?: number\n /** Deterministic seed for a reproducible design. */\n seed?: number\n /** Design quality knob (provider-specific). */\n quality?: number\n /** Enhance the generated voice. */\n shouldEnhance?: boolean\n /** Optional extra prompt context (≤8000 chars). */\n userPrompt?: string\n}\n\n/** Input for {@link VoicesResource.remix}. */\nexport interface VoiceRemixInput {\n /** The text (1–5000 chars) to speak in the described voice. */\n text: string\n /** Natural-language description of the voice to speak in. */\n voiceDescription: string\n /** Optional extra prompt context (≤8000 chars). */\n userPrompt?: string\n}\n\n/** Input for {@link VoicesResource.dub}. Provide exactly ONE source:\n * `audioUrl` (dubbed audio out), `videoUrl` (the dubbed VIDEO out, plus the\n * dubbed audio track as `output_data.audioUrl`), or `sourceUrl` (a link to\n * a video's page, or a direct link to a file, that the dubbing service\n * fetches itself). Priced per minute of the dubbed span; the span is capped\n * at 30 minutes — use `startTime`/`endTime` to dub part of a longer source. */\nexport interface DubbingInput {\n /** URL of the audio to dub (audio mode). */\n audioUrl?: string\n /** URL of the video to dub — the result is the dubbed VIDEO (+ audio track). */\n videoUrl?: string\n /**\n * A link to a video's page, or a direct link to a file, that the dubbing\n * service fetches itself: an import of that link, made under the content\n * rights declaration (see `confirmationId`).\n */\n sourceUrl?: string\n /**\n * The content rights confirmation for `sourceUrl` (`contentRights.confirm()`).\n * Without one, a deployment that requires a confirmation refuses the dub with\n * `403`, unless the request comes through a key or an agent whose account\n * holds a standing authorization.\n */\n confirmationId?: string\n /** Target language ISO code (2–10 chars), e.g. `\"es\"`, `\"pt-BR\"`. */\n targetLanguage: string\n /** Source language ISO code; auto-detected when omitted. */\n sourceLanguage?: string\n /** Expected number of speakers (1–20; 0 = auto-detect) — improves separation when known. */\n numSpeakers?: number\n /** Keep the original voices instead of cloning them into the target language. */\n disableVoiceCloning?: boolean\n /** Drop the background/music bed from the dubbed output. */\n dropBackgroundAudio?: boolean\n /** Dub only this window of the source (seconds). */\n startTime?: number\n endTime?: number\n /** Keep the source resolution on video dubs (slower render). */\n highestResolution?: boolean\n useProfanityFilter?: boolean\n /** Experimental: steer dubbed voices toward an accent. */\n targetAccent?: string\n /** Apply ElevenLabs' own watermark to video dubs. */\n watermark?: boolean\n}\n"]}
@@ -28,7 +28,9 @@ var CastSwapResource = class {
28
28
  * nothing to swap restyles the whole clip there. Where it is not `true`
29
29
  * (absent, or an engine missing from it, included) on Nodaro Motion Prism, an
30
30
  * art style alone is refused, and one with a pre or post prompt runs as any
31
- * other art-style run.
31
+ * other art-style run. Its `preciseArtStyle` says, per engine, whether a
32
+ * Precise run takes an art style there too, and its `preciseArtStyleAlone`
33
+ * whether that art style alone is the run's change.
32
34
  * Public (no sign-in needed); the answer may be up to a minute old.
33
35
  */
34
36
  styleChoice() {
@@ -72,5 +74,5 @@ var CastSwapResource = class {
72
74
  };
73
75
 
74
76
  export { CastSwapResource };
75
- //# sourceMappingURL=chunk-ZAATTX5W.js.map
76
- //# sourceMappingURL=chunk-ZAATTX5W.js.map
77
+ //# sourceMappingURL=chunk-DHJRS65Q.js.map
78
+ //# sourceMappingURL=chunk-DHJRS65Q.js.map