@nodaro/sdk 2.29.0 → 2.32.0

package/dist/entries/voices.d.cts

@@ -0,0 +1,498 @@
+import { Voice, VoiceLibraryParams, VoiceLibraryResponse, VoiceClone } from '@nodaro/shared/pipeline';
+export { SharedVoice, Voice, VoiceClone, VoiceLibraryParams, VoiceLibraryResponse } from '@nodaro/shared/pipeline';
+import { AudioFxPreset } from '@nodaro/shared/media';
+export { AudioFxPreset } from '@nodaro/shared/media';
+import { DialogueProvider } from '@nodaro/shared/models';
+import { N as NodaroClientCore } from '../developer-apps-zSQXWmEG.cjs';
+import '@nodaro/shared/apps';
+import '@nodaro/shared/organizations';
+
+/**
+ * Re-export the shared voice types so SDK consumers don't have to add
+ * `@nodaro/shared` as a second dependency just to type a `Voice` row, a
+ * `VoiceClone`, or a `searchLibrary` call. Single source of truth lives in
+ * `@nodaro/shared`.
+ */
+/** Audio-FX preset union (reverb spaces / telephone / megaphone / echo / custom) — used by {@link VoiceChangerProInput.voiceFx}. */
+/**
+ * Read access to ElevenLabs voices: the premade catalog plus the shared
+ * community Voice Library (both public GETs, no body), and the signed-in
+ * user's own pre-retirement voice clones (list / delete).
+ */
+declare class VoicesResource {
+    private client;
+    constructor(client: NodaroClientCore);
+    /**
+     * List the premade ElevenLabs voices (`GET /v1/voices`). Falls back to a
+     * curated set server-side when no ElevenLabs API key is configured.
+     */
+    list(): Promise<Voice[]>;
+    /**
+     * Search the shared/community Voice Library (`GET /v1/voices/library`). All
+     * params are optional and forwarded as a querystring; `undefined` / `null` /
+     * empty-string values are omitted so the server defaults apply. `hasMore`
+     * drives "load more" pagination.
+     */
+    searchLibrary(params?: VoiceLibraryParams): Promise<VoiceLibraryResponse>;
+    /**
+     * List the signed-in user's voice clones (`GET /v1/voice-clones`). The route
+     * wraps the rows in `{ voiceClones }`; we unwrap to the bare array.
+     */
+    listClones(): Promise<VoiceClone[]>;
+    /**
+     * @deprecated Voice cloning is no longer offered on Nodaro (retired
+     * 2026-09-15). The route answers `410 voice_cloning_retired` and this call
+     * rejects with that error. Clones created before the retirement are still
+     * listed by {@link VoicesResource.listClones} and still work as voice ids.
+     * For a new custom voice use {@link VoicesResource.design}.
+     */
+    createClone(input: {
+        name: string;
+        audioUrl: string;
+    }): Promise<VoiceClone>;
+    /** Delete one of the user's voice clones (`DELETE /v1/voice-clones/:id`). */
+    deleteClone(id: string): Promise<void>;
+    /**
+     * Replace the voice in a recording — or in a whole talking video — with a
+     * different voice (`POST /v1/voice-changer`). Pass `audioUrl` to revoice
+     * audio→audio, or `videoUrl` to revoice an entire clip (the server demuxes
+     * the audio, runs speech-to-speech, and remuxes onto the original video,
+     * returning the video plus the new audio track). Exactly one of `audioUrl` /
+     * `videoUrl` is required; when both are sent, video wins. `removeBackgroundNoise`
+     * off keeps the music/SFX bed under the new voice; on yields a clean voice-only
+     * result. Costs credits and runs async — poll `jobs.get(jobId)` for the result
+     * (`output_data.videoUrl` + `output_data.audioUrl` in video mode).
+     */
+    change(input: {
+        voiceId: string;
+        audioUrl?: string;
+        videoUrl?: string;
+        /** Speech-to-speech model id. Defaults to the server-configured default when omitted. */
+        model?: string;
+        stability?: number;
+        similarityBoost?: number;
+        /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency/stability. */
+        style?: number;
+        /** ElevenLabs speaker boost — sharpens fidelity to the target speaker (small latency cost). */
+        useSpeakerBoost?: boolean;
+        /** Deterministic speech-to-speech seed (integer 0–4294967295) for reproducible output. Omit for random. */
+        seed?: number;
+        removeBackgroundNoise?: boolean;
+    }): Promise<{
+        jobId: string;
+    }>;
+    /**
+     * Recast each detected speaker in a multi-speaker recording to a different
+     * voice (`POST /v1/voice-changer-pro`). `orderedVoices` maps speaker positions to
+     * voices in detection order — speaker 0 → `orderedVoices[0]`, speaker 1 →
+     * `orderedVoices[1]`, etc. Speakers beyond the end of `orderedVoices` keep
+     * their original voice. Each entry is EITHER a bare voice id (premade name or
+     * ElevenLabs UUID) OR a {@link VoiceChangerProVoice} object carrying per-voice
+     * ElevenLabs speech-to-speech settings (stability / similarityBoost / style /
+     * useSpeakerBoost / `seed`) plus a loudness `volumeMode` (and a manual
+     * `volume`) — OR `null`, meaning keep this speaker's original voice (requires
+     * a platform running cloud-plugins with keep-slot support). At least one
+     * entry must be non-null. A per-voice `seed` makes that speaker's recast
+     * reproducible.
+     *
+     * Pass `audioUrl` for audio-only recast or `videoUrl` to recast the audio
+     * track of a video clip (the server demuxes, recasts, and remuxes).
+     *
+     * Voice and music are ALWAYS separated first — before recasting, the source
+     * is split into an isolated vocal stem and a music/SFX stem.
+     * `preserveBackground` (default `true`) only controls whether that
+     * music/instrumental stem is mixed back under the new voices; set it `false`
+     * for a clean voice-only result. `separationQuality` selects the quality of
+     * the voice/music separation: `"fast"` (default, quicker — preserves more of
+     * the voice) or `"best"` (finer voice/music separation).
+     * `removeBackgroundNoise` additionally denoises the result.
+     * `musicVolumeMode` sets the level of that preserved background (only relevant
+     * when `preserveBackground` is on): `"match"` (default) keeps the original
+     * level, `"normalize"` loudnorms it, `"manual"` uses `musicVolume`%.
+     * `voiceFx` applies a reverb/echo to the COMBINED recast voices BEFORE the
+     * background is mixed back in (effect sits on the voices, not the music bed).
+     *
+     * Cloud-only — costs credits and runs async; poll `jobs.get(jobId)` for the
+     * result (`output_data.videoUrl` + `output_data.audioUrl` in video mode).
+     */
+    recast(input: VoiceChangerProInput): Promise<{
+        jobId: string;
+    }>;
+    /**
+     * Detect the speakers in a clip WITHOUT recasting yet
+     * (`POST /v1/voice-changer-pro/analyze`) — the first step of the interactive
+     * flow. Separates voice from music once and diarizes the vocals, returning the
+     * speaker list so a user (or agent) can choose a voice per speaker before
+     * committing to a paid recast. Poll `jobs.get(jobId)`: the completed job's
+     * `output_data` carries the separated stem urls + the detected `speakers`
+     * (each with `id`, time `segments`, `firstStartSec`, `wordCount`, `snippet`)
+     * and the detected language — reshape it into a {@link VcpAnalysis} and pass it
+     * as `recast({ ..., analysis })` to skip re-detection. With `suggestTitle`,
+     * `output_data.suggestedTitle` also carries an LLM-proposed title.
+     *
+     * Cloud-only; costs credits and runs async.
+     */
+    analyze(input: VcpAnalyzeInput): Promise<{
+        jobId: string;
+    }>;
+    /**
+     * Render a final video from a mixed set of stems
+     * (`POST /v1/voice-changer-pro/export`) — the last step of the interactive
+     * flow. After `recast({ output: "stems" })` hands back the dry per-track stems
+     * and the user has set levels / mutes / an effect in your editor, pass those
+     * `tracks` (plus the source `videoUrl`) here to mix and remux into the finished
+     * video. The video is stream-copied (never re-encoded), so the export is
+     * bit-identical to your preview. At least one track must be un-muted (all-muted
+     * is a 400); `voiceFx` is applied to the voice tracks at render time.
+     *
+     * Cloud-only; costs credits and runs async — poll `jobs.get(jobId)` for the
+     * result (`output_data.videoUrl`).
+     */
+    exportMix(input: VcpExportInput): Promise<{
+        jobId: string;
+    }>;
+    /**
+     * Design a brand-new synthetic voice from a text description
+     * (`POST /v1/voice-design`) — ElevenLabs text-to-voice. `text` (100–1000 chars)
+     * is a preview line spoken in the designed voice; `voiceDescription` describes
+     * the voice to create. Costs credits and runs async — poll `jobs.get(jobId)`
+     * for the preview + the reusable voice id.
+     */
+    design(input: VoiceDesignInput): Promise<{
+        jobId: string;
+    }>;
+    /**
+     * Generate speech in a voice described in natural language, without cloning
+     * (`POST /v1/voice-remix`). `text` (1–5000 chars) is spoken in a voice matching
+     * `voiceDescription`. Costs credits and runs async — poll `jobs.get(jobId)`.
+     */
+    remix(input: VoiceRemixInput): Promise<{
+        jobId: string;
+    }>;
+    /**
+     * Dub an audio clip into another language while preserving each speaker's voice
+     * (`POST /v1/dubbing`). `targetLanguage` is an ISO code (e.g. `"es"`, `"fr"`);
+     * `sourceLanguage` is auto-detected when omitted. Costs credits and runs async
+     * — poll `jobs.get(jobId)`.
+     */
+    dub(input: DubbingInput): Promise<{
+        jobId: string;
+    }>;
+    /**
+     * Voice a multi-speaker script as ONE audio file
+     * (`POST /v1/text-to-dialogue`) on ElevenLabs Dialogue v3 — the default — or
+     * Dialogue v4 (`provider: "elevenlabs-dialogue-v4"`). Each `dialogue` line is
+     * `{ text, voice }` in speaking order — `voice` is a premade voice name or an
+     * ElevenLabs voice UUID (cloned/library voices work too; mixed casts are
+     * fine). At most 5,000 characters total across lines on Dialogue v3 and
+     * 10,000 on Dialogue v4 (≤2,000 recommended for best quality) and 10 unique
+     * voices. Line text may
+     * carry `[audio tags]` like `[laughs]`. Costs credits and runs async — poll
+     * `jobs.get(jobId)` for `output_data.audioUrl` and — on every dialogue model,
+     * since both return timings at no extra credits — `output_data.transcript`, a
+     * `Transcript` (`@nodaro/shared`): `words[]` with `startMs` / `endMs` /
+     * `speaker` and one `segments[]` entry per line, `speaker` being the line's
+     * `voice` as you sent it. Pass `transcript.words` to
+     * `addCaptions({ captions, autoTranscribe: false })` to caption the dialogue
+     * without a second transcription.
+     */
+    textToDialogue(input: {
+        dialogue: Array<{
+            text: string;
+            voice: string;
+        }>;
+        /** Dialogue model: `"elevenlabs-dialogue"` (v3, the default) or `"elevenlabs-dialogue-v4"`. */
+        provider?: DialogueProvider;
+        /** v3: exactly 0, 0.5 or 1 (0 = most variable, 1 = most stable). v4: any value 0–1. */
+        stability?: number;
+        /** v4 only (v3 ignores it): how closely each line keeps its voice's character, 0–1. */
+        similarityBoost?: number;
+        /** ISO 639-1 language hint (e.g. "en"); auto-detected when omitted. */
+        languageCode?: string;
+        /** Deterministic sampling seed (integer 0–4294967295). Omit for random. */
+        seed?: number;
+        /** Spell out numbers/dates/abbreviations: "auto" (default) | "on" | "off". */
+        applyTextNormalization?: "auto" | "on" | "off";
+    }): Promise<{
+        jobId: string;
+    }>;
+    /**
+     * @deprecated Voice cloning is no longer offered on Nodaro (retired
+     * 2026-09-15). The route answers `410 voice_cloning_retired` and this call
+     * rejects with that error. See {@link VoicesResource.createClone}.
+     */
+    createCloneFromFile(input: {
+        name: string;
+        file: Blob | Uint8Array | ArrayBuffer;
+        /** File name for the upload part (default `sample`). */
+        filename?: string;
+        /** MIME type when `file` is a raw buffer (default `audio/mpeg`). */
+        contentType?: string;
+    }): Promise<VoiceClone>;
+}
+/**
+ * One entry in {@link VoiceChangerProInput.orderedVoices}. Either a bare voice id
+ * (premade name like `"Rachel"` or an ElevenLabs UUID for a custom clone), or
+ * an object pinning per-voice speech-to-speech settings and the recast's
+ * loudness behaviour for that speaker.
+ */
+type VoiceChangerProVoice = string | {
+    /** Target voice — premade name (`"Rachel"`, `"Aria"`, …) or an ElevenLabs UUID for a custom clone. */
+    voiceId: string;
+    /**
+     * Which lane converts this speaker. `"sts"` (default) is the classic
+     * speech-to-speech recast; `"v3"` and `"v4"` are RE-SPEAK — the
+     * performance is REGENERATED from the transcript (supports `[audio
+     * tags]`; original delivery is replaced, and lips won't match on video).
+     * A Re-speak speaker needs transcript text: pass an `analysis` whose
+     * `segments[].text` carries it (analyze now emits this), or omit
+     * `analysis` and the engine re-speaks from its own transcription. For
+     * `"v3"` (eleven_v3), `stability` accepts exactly 0 / 0.5 / 1, and
+     * `similarityBoost` / `style` / `useSpeakerBoost` are ignored. For
+     * `"v4"` (eleven_v4), `stability` takes any value 0–1 and
+     * `similarityBoost` is honoured (`style` / `useSpeakerBoost` ignored);
+     * each line is generated with its neighbouring lines as context for
+     * smoother joins. Both Re-speak engines are priced the same, per started
+     * 1,000 characters.
+     */
+    engine?: "sts" | "v3" | "v4";
+    /** ElevenLabs stability (0–1). Higher = steadier, lower = more expressive. */
+    stability?: number;
+    /** ElevenLabs similarity boost (0–1) — how closely the output hugs the target voice's timbre. */
+    similarityBoost?: number;
+    /** Style exaggeration (0–1). Default 0; >0 amplifies delivery at the cost of latency / stability. */
+    style?: number;
+    /** ElevenLabs speaker boost — sharpens fidelity to the target speaker. */
+    useSpeakerBoost?: boolean;
+    /**
+     * Deterministic speech-to-speech seed (integer 0–4294967295) for
+     * reproducible per-voice output — the same source + settings + seed
+     * recast this speaker identically across runs. Omit for a random seed.
+     */
+    seed?: number;
+    /**
+     * Loudness handling for this recast voice. `"match"` (default) matches the
+     * original speaker's loudness; `"normalize"` applies loudness
+     * normalization; `"manual"` uses `volume` as a percentage.
+     */
+    volumeMode?: "match" | "normalize" | "manual";
+    /** Manual output volume as a percentage (0–200). Consulted only when `volumeMode === "manual"`. */
+    volume?: number;
+};
+/** Input for {@link VoicesResource.recast}. */
+interface VoiceChangerProInput {
+    /** URL of an audio file to recast (audio → audio). Exactly one of `audioUrl` / `videoUrl` is required. */
+    audioUrl?: string;
+    /** URL of a video file to recast (the audio track is recast and remuxed). Exactly one of `audioUrl` / `videoUrl` is required. */
+    videoUrl?: string;
+    /**
+     * Voices in speaker-detection order. Speaker N is mapped to `orderedVoices[N]`;
+     * speakers beyond the array keep their original voice. Each entry is a bare
+     * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings —
+     * OR `null`, meaning keep this speaker's original voice (requires a platform
+     * running cloud-plugins with keep-slot support). At least one entry must be
+     * non-null.
+     */
+    orderedVoices: Array<VoiceChangerProVoice | null>;
+    /** Model to use for speech-to-speech. Defaults to the server-configured default when omitted. */
+    model?: string;
+    /**
+     * Mix the separated music / SFX stem back under the recast voices. Default
+     * `true`. The voice is ALWAYS split out before recasting regardless of this
+     * flag — `false` simply drops the music for a clean voice-only result.
+     */
+    preserveBackground?: boolean;
+    /**
+     * Quality of the voice/music separation. `"fast"` (default, quicker —
+     * preserves more of the voice) or `"best"` (finer voice/music separation).
+     */
+    separationQuality?: "fast" | "best";
+    /** Strip background noise for a clean voice-only result. */
+    removeBackgroundNoise?: boolean;
+    /**
+     * Level of the preserved background music / SFX stem in the final mix. Only
+     * relevant when `preserveBackground` is on (otherwise there is no background to
+     * level). `"match"` (default) leaves the separated instrumental at its original
+     * level; `"normalize"` applies loudness normalization; `"manual"` sets its
+     * level to `musicVolume`%.
+     */
+    musicVolumeMode?: "match" | "normalize" | "manual";
+    /** Background music level as a percentage (0–200). Consulted only when `musicVolumeMode === "manual"`. */
+    musicVolume?: number;
+    /**
+     * Node-level reverb/echo applied to the COMBINED recast voices **before** the
+     * background is mixed back in (so the effect sits on the voices only, not the
+     * music/SFX bed). Reverb presets (`room`, `hall`, `church`, …) use
+     * `wetDryMix`; the `echo` / `custom` presets use `delayMs` + `decay`. Omit for
+     * no effect.
+     */
+    voiceFx?: {
+        /** Effect preset — reverb space, `telephone`, `megaphone`, `echo`, or `custom`. */
+        preset: AudioFxPreset;
+        /** Reverb wet/dry mix as a percentage (0–100). Higher = wetter (more reverb). */
+        wetDryMix?: number;
+        /** Echo delay in milliseconds (20–2000). Used by the `echo` / `custom` presets. */
+        delayMs?: number;
+        /** Echo decay / feedback (0–1). Higher = more repeats. Used by the `echo` / `custom` presets. */
+        decay?: number;
+    };
+    /**
+     * Output mode. `"video"` (default) mixes the recast voices with the preserved
+     * background and returns a finished merged video. `"stems"` returns the dry,
+     * unleveled per-track stems instead (rendering nothing) so you can drive an
+     * INTERACTIVE mix — adjust levels/mutes/effect in your own UI, then render the
+     * final video with {@link VoicesResource.exportMix}. This is how an app builds
+     * a full editor around VCP rather than a one-shot recast.
+     */
+    output?: "video" | "stems";
+    /**
+     * A prior {@link VoicesResource.analyze} result. Pass it to SKIP re-detection:
+     * the recast reuses the already-separated stems and speaker segments instead of
+     * running separation + diarization again. This is the fast-path for the
+     * detect → pick voices → recast interactive flow (analyze once, recast N times
+     * as the user tweaks voice assignments). Omit to auto-detect from the source.
+     */
+    analysis?: VcpAnalysis;
+}
+/** One detected speaker in a {@link VcpAnalysis} (from `analyze`). */
+interface VcpAnalysisSpeaker {
+    /** Stable speaker id (first-appearance order). */
+    id: string;
+    /** The speaker's spoken time ranges (seconds). `text` is what was said in
+     *  the range — the paid input for a Re-speak speaker (`engine: "v3"` or
+     *  `"v4"`; editable before conversion); the STS lane ignores it. */
+    segments: Array<{
+        start: number;
+        end: number;
+        text?: string;
+    }>;
+    /** When the speaker first speaks (seconds). */
+    firstStartSec?: number;
+    /** Rough word count across the clip — a proxy for how much this speaker says. */
+    wordCount?: number;
+    /** The first few transcribed words, to help a user tell speakers apart. */
+    snippet?: string;
+}
+/**
+ * The result of {@link VoicesResource.analyze}, reshaped to pass back into
+ * {@link VoiceChangerProInput.analysis}. Read a completed analyze job's
+ * `output_data` into this shape (it carries the separated stem urls + the
+ * detected speakers) and thread it into `recast` to skip re-detection.
+ */
+interface VcpAnalysis {
+    /** URL of the isolated vocal stem. */
+    vocalsUrl: string;
+    /** URL of the separated music/SFX stem (absent when the source had none). */
+    backgroundUrl?: string;
+    /** The detected speakers, in first-appearance order. */
+    speakers: VcpAnalysisSpeaker[];
+    /** Scribe's detected language code, round-tripped so the recast auto-selects the STS model. */
+    languageCode?: string;
+    /** Confidence (0–1) of {@link VcpAnalysis.languageCode}. */
+    languageProbability?: number;
+}
+/** Input for {@link VoicesResource.analyze}. */
+interface VcpAnalyzeInput {
+    /** URL of an audio file to analyze. Exactly one of `audioUrl` / `videoUrl` is required. */
+    audioUrl?: string;
+    /** URL of a video file to analyze (its audio track is used). Exactly one of `audioUrl` / `videoUrl` is required. */
+    videoUrl?: string;
+    /** Quality of the voice/music separation run before diarization: `"fast"` (default) or `"best"`. */
+    separationQuality?: "fast" | "best";
+    /** Also suggest a conversion title from the transcript (returned on the job's `output_data.suggestedTitle`). */
+    suggestTitle?: boolean;
+}
+/** One track in a {@link VcpExportInput} mix. */
+interface VcpExportTrack {
+    /** URL of the stem for this lane (a recast voice stem or the background stem). */
+    url: string;
+    /** Fader position as a percentage: 0 = silent, 100 = unity, 200 = +6dB. */
+    gain: number;
+    /** Whether this lane is muted in the mix. */
+    muted: boolean;
+    /**
+     * Which bucket the track is in, and so whether `voiceFx` lands on it. Defaults
+     * to `"voice"`. Set `"background"` for the music/SFX lane (the effect never
+     * touches it).
+     */
+    kind?: "voice" | "background";
+}
+/** Input for {@link VoicesResource.exportMix}. */
+interface VcpExportInput {
+    /** The source video to remux the mixed audio onto (stream-copied — never re-encoded). */
+    videoUrl: string;
+    /** The mix: one entry per lane. At least one must be un-muted (all-muted is a 400). Max 16 tracks. */
+    tracks: VcpExportTrack[];
+    /**
+     * A reverb/echo applied to the VOICE tracks only (not `"background"` lanes)
+     * at render time — so iterating the effect in your editor is free until you
+     * export. Same shape as {@link VoiceChangerProInput.voiceFx}.
+     */
+    voiceFx?: VoiceChangerProInput["voiceFx"];
+}
+/** Input for {@link VoicesResource.design}. */
+interface VoiceDesignInput {
+    /** A preview line (100–1000 chars) spoken in the designed voice. */
+    text: string;
+    /** Natural-language description of the voice to create. */
+    voiceDescription: string;
+    /** Voice-design model id. Defaults to the server-configured default. */
+    model?: string;
+    /** Output loudness (-1..1). */
+    loudness?: number;
+    /** How strongly the description steers the design (0–100). */
+    guidanceScale?: number;
+    /** Deterministic seed for a reproducible design. */
+    seed?: number;
+    /** Design quality knob (provider-specific). */
+    quality?: number;
+    /** Enhance the generated voice. */
+    shouldEnhance?: boolean;
+    /** Optional extra prompt context (≤8000 chars). */
+    userPrompt?: string;
+}
+/** Input for {@link VoicesResource.remix}. */
+interface VoiceRemixInput {
+    /** The text (1–5000 chars) to speak in the described voice. */
+    text: string;
+    /** Natural-language description of the voice to speak in. */
+    voiceDescription: string;
+    /** Optional extra prompt context (≤8000 chars). */
+    userPrompt?: string;
+}
+/** Input for {@link VoicesResource.dub}. Provide exactly ONE source:
+ *  `audioUrl` (dubbed audio out), `videoUrl` (the dubbed VIDEO out, plus the
+ *  dubbed audio track as `output_data.audioUrl`), or `sourceUrl` (a public
+ *  YouTube/TikTok/direct link ElevenLabs fetches itself). Priced per minute
+ *  of the dubbed span; the span is capped at 30 minutes — use
+ *  `startTime`/`endTime` to dub part of a longer source. */
+interface DubbingInput {
+    /** URL of the audio to dub (audio mode). */
+    audioUrl?: string;
+    /** URL of the video to dub — the result is the dubbed VIDEO (+ audio track). */
+    videoUrl?: string;
+    /** Public page/media URL (YouTube, TikTok, or a direct link) — ElevenLabs fetches it directly. */
+    sourceUrl?: string;
+    /** Target language ISO code (2–10 chars), e.g. `"es"`, `"pt-BR"`. */
+    targetLanguage: string;
+    /** Source language ISO code; auto-detected when omitted. */
+    sourceLanguage?: string;
+    /** Expected number of speakers (1–20; 0 = auto-detect) — improves separation when known. */
+    numSpeakers?: number;
+    /** Keep the original voices instead of cloning them into the target language. */
+    disableVoiceCloning?: boolean;
+    /** Drop the background/music bed from the dubbed output. */
+    dropBackgroundAudio?: boolean;
+    /** Dub only this window of the source (seconds). */
+    startTime?: number;
+    endTime?: number;
+    /** Keep the source resolution on video dubs (slower render). */
+    highestResolution?: boolean;
+    useProfanityFilter?: boolean;
+    /** Experimental: steer dubbed voices toward an accent. */
+    targetAccent?: string;
+    /** Apply ElevenLabs' own watermark to video dubs. */
+    watermark?: boolean;
+}
+
+export { type DubbingInput, type VcpAnalysis, type VcpAnalysisSpeaker, type VcpAnalyzeInput, type VcpExportInput, type VcpExportTrack, type VoiceChangerProInput, type VoiceChangerProVoice, type VoiceDesignInput, type VoiceRemixInput, VoicesResource };