@nodaro/sdk 2.12.0 → 2.16.0

package/dist/index.cjs

@@ -2156,9 +2156,13 @@ var MediaResource = class {
    * Download a social video (YouTube / TikTok / Instagram / X / Facebook) into
    * your storage (`POST /v1/download-video`). `maxHeight` caps the resolution
    * (default "best"); `sectionStartSec` + `sectionEndSec` (both-or-neither) fetch
-   * ONLY that time range instead of the whole video. Returns a `downloadId`;
-   * progress streams from `GET /v1/download-video/progress/:downloadId`
-   * (server-sent events) and the finished file lands in your library.
+   * ONLY that time range instead of the whole video. A download that arrives
+   * with no audio stream FAILS by default (it is usually a degraded source
+   * response, and is retried through other routes first); pass
+   * `requireAudio: false` to accept a clip that really has no sound. Returns a
+   * `downloadId`; progress streams from
+   * `GET /v1/download-video/progress/:downloadId` (server-sent events) and the
+   * finished file lands in your library.
    */
   downloadVideo(input) {
     return this.client.request("POST", "/v1/download-video", { body: input });
@@ -2260,6 +2264,52 @@ var MediaResource = class {
   trimVideo(input) {
     return this.client.request("POST", "/v1/trim-video", { body: input });
   }
+  /**
+   * Burn captions into a video (`POST /v1/add-captions`). Give the words as
+   * `text`, word-timed `captions[]`, or let it transcribe (`autoTranscribe`,
+   * the default when neither is set).
+   *
+   * `style: "subtitle"` renders statically (FFmpeg) UNLESS it carries a styling
+   * lever, in which case it — like the KINETIC styles (`word-highlight` /
+   * `karaoke` / `tiktok-words` / `word-pop` / `bouncy`) — renders via Remotion.
+   * `look` picks a preset — `outline` (Montserrat 900, UPPERCASE, black outline,
+   * yellow spoken word — the TikTok/Reels read) or `clean`; an UNSET look renders
+   * as `outline`. The explicit STYLING levers (`fontFamily`, `fontWeight`,
+   * `strokeColor`/`strokeWidth`, `uppercase`, `positionY`, `maxWordsPerLine`)
+   * override individual fields of it and now apply to `subtitle` too. Only
+   * `highlightColor` (the spoken-word cursor) and `animate` stay kinetic-only
+   * and are REJECTED on `subtitle`.
+   *
+   * `animate` (default true) freezes the per-word MOTION on the kinetic styles
+   * when set to false — the grouping, line-holding and spoken-word highlight
+   * stay; only the movement stops.
+   *
+   * `maxWordsPerLine` (1-20) caps how many words a line — or a `tiktok-words`
+   * page — may hold, on top of the width budget / sentence ends / pauses that
+   * already close one; it is inert on `word-pop`, which is always one word.
+   * Like every lever here it exists top-level AND per segment, and a segment
+   * that does not name its own inherits the top-level value.
+   *
+   * `segments[]` applies DIFFERENT treatments to non-overlapping time ranges in
+   * one call (e.g. a large top intro, then a small bottom body); a segment that
+   * names its own `look` starts fresh from that preset and does not inherit the
+   * top-level explicit levers. Poll `jobs.get(jobId)`.
+   *
+   * `word-highlight` shows ONE held line at a time and a word's
+   * `startMs`/`endMs` is its SPOKEN window (what times the highlight, not how
+   * long the text is on screen) — so `captions[]` can be handed over verbatim
+   * from an `audio.transcribe()` job's `output_data.words`, with
+   * `autoTranscribe: false`.
+   */
+  addCaptions(input) {
+    const { autoTranscribe, transcribeProvider, ...rest } = input;
+    const body = {
+      ...rest,
+      ...autoTranscribe !== void 0 ? { auto_transcribe: autoTranscribe } : {},
+      ...transcribeProvider !== void 0 ? { transcribe_provider: transcribeProvider } : {}
+    };
+    return this.client.request("POST", "/v1/add-captions", { body });
+  }
   /**
    * Trim (and extract) audio from a video or audio source
    * (`POST /v1/trim-audio`) to `[startTime, endTime]` seconds, in `audioFormat`
@@ -2292,6 +2342,25 @@ var MediaResource = class {
   slideshow(input) {
     return this.client.request("POST", "/v1/slideshow", { body: input });
   }
+  /**
+   * Place 1–20 timed image layers over a video (`POST /v1/video-overlay`) —
+   * rendered locally (FFmpeg) in one pass, no AI; the base audio is kept
+   * untouched. Each layer has an `imageUrl`, a `start` and an optional `end`
+   * in seconds (no `end` = to the end of the video) and a placement: a
+   * `preset` (`"card"`, `"corner-badge"` with a `corner`, `"full-frame"`) or
+   * an explicit box in {@link MediaResource.imageOverlay}'s percent vocabulary
+   * (`anchor`, `x`, `y`, `width`, `height`, `fit`) — an explicit box field
+   * overrides the preset, and a layer with neither is a corner badge
+   * (bottom-right, or the `corner` it names). `outputAspect` renders onto a
+   * 16:9 / 9:16 / 1:1 / 4:5 canvas
+   * (`baseFit` cover by default; `backgroundColor` pads `contain`). Poll
+   * `jobs.get(jobId)`: the output carries `videoUrl`, `thumbnailUrl`, `width`,
+   * `height`, `durationSec` and `warnings[]` (a layer clipped or skipped at the
+   * video's end, an animated image's first frame, re-encoded audio).
+   */
+  videoOverlay(input) {
+    return this.client.request("POST", "/v1/video-overlay", { body: input });
+  }
   /**
    * Probe a social video's metadata (`POST /v1/video-metadata`) — duration,
    * dimensions, title, live status — WITHOUT downloading it. A direct read, not a
@@ -2363,6 +2432,32 @@ var AudioResource = class {
   combine(input) {
     return this.client.request("POST", "/v1/combine-audio", { body: input });
   }
+  /**
+   * Transcribe an audio (or video) track to text (`POST /v1/transcribe`).
+   *
+   * Three engines are accepted. `elevenlabs-stt` (Scribe) is always word-level,
+   * flag or not, and is the only lane that honours `diarize` (who spoke) and
+   * `tagAudioEvents` (laughter, applause, …); `incredibly-fast-whisper` returns
+   * word timings when you ask for them. `whisper` returns NO word timings at
+   * all — named explicitly or reached by OMITTING `provider`, which still falls
+   * back to it — so asking it for them (`wordTimestamps: true`) is rejected with
+   * a `400 validation_error` at ingress, before any credit is spent. A kinetic
+   * caption render therefore has to name one of the other two.
+   *
+   * Poll `jobs.get(jobId)`; the finished job's `output_data` is a
+   * {@link TranscribeJobOutput}: `text` (the whole transcript), `words`
+   * (caption-shaped, in MILLISECONDS), `json` (the normalized
+   * {@link Transcript}, also ms) — and a top-level `segments` array that is in
+   * SECONDS, not ms.
+   *
+   * `words` is the caption source for a kinetic burn-in: hand it to
+   * `media.addCaptions()` as `captions` with `autoTranscribe: false` and the
+   * render uses those exact words (correct the `text` of an entry in between
+   * and the fix is what burns in).
+   */
+  transcribe(input) {
+    return this.client.request("POST", "/v1/transcribe", { body: input });
+  }
 };
 
 // src/resources/credits.ts
@@ -3451,7 +3546,104 @@ var WorkspacesResource = class {
     return this.client.requestText("GET", `/v1/workspaces/${encodeURIComponent(id)}/usage`, { query: { ...opts, format: "csv" } });
   }
 };
-var SDK_VERSION = "2.12.0" ;
+var EditResource = class {
+  constructor(client) {
+    this.client = client;
+  }
+  client;
+  /**
+   * Detect silence ranges in an audio/video source (`POST /v1/silence-detect`).
+   * Keyless — one ffmpeg pass over the source's audio proxy.
+   */
+  silenceDetect(input) {
+    return this.client.request("POST", "/v1/silence-detect", {
+      body: {
+        audioUrl: input.audioUrl,
+        ...input.thresholdDb !== void 0 ? { thresholdDb: input.thresholdDb } : {},
+        ...input.minSilenceMs !== void 0 ? { minSilenceMs: input.minSilenceMs } : {},
+        ...input.padMs !== void 0 ? { padMs: input.padMs } : {},
+        ...input.workflowId !== void 0 ? { workflowId: input.workflowId } : {}
+      }
+    });
+  }
+  /**
+   * Measure how far apart the clocks of 2–6 recordings of one conversation are
+   * (`POST /v1/audio-sync`), by cross-correlating their audio — keyless, one
+   * ffmpeg decode per source. The finished job's `output_data.json` is an
+   * {@link AudioSyncResult}. A malformed request (fewer than 2 or more than 6
+   * sources, a repeated id, a `reference` that is not one of the ids) throws a
+   * typed `NodaroError` (400, `code: "validation_error"`) before any credits
+   * are reserved.
+   */
+  audioSync(input) {
+    return this.client.request("POST", "/v1/audio-sync", {
+      body: {
+        sources: input.sources.map((s) => ({ id: s.id, url: s.url })),
+        ...input.reference !== void 0 ? { reference: input.reference } : {},
+        ...input.workflowId !== void 0 ? { workflowId: input.workflowId } : {}
+      }
+    });
+  }
+  /**
+   * Render an edit decision list into a video or audio cut (`POST /v1/apply-edl`).
+   * The EDL is validated at ingress — an unresolvable source or a picture-less
+   * segment on a video edit throws a typed `NodaroError` (400, `code:
+   * "invalid_edl"`) before any credits are reserved.
+   */
+  applyEdl(input) {
+    return this.client.request("POST", "/v1/apply-edl", {
+      body: {
+        edl: input.edl,
+        ...input.sources !== void 0 ? { sources: input.sources } : {},
+        ...input.transcript !== void 0 ? { transcript: input.transcript } : {},
+        ...input.output !== void 0 ? { output: input.output } : {},
+        ...input.quality !== void 0 ? { quality: input.quality } : {},
+        ...input.crossfadeMs !== void 0 ? { crossfadeMs: input.crossfadeMs } : {},
+        ...input.workflowId !== void 0 ? { workflowId: input.workflowId } : {}
+      }
+    });
+  }
+  /**
+   * Plan a transcript-driven cut (`POST /v1/edit-plan`). Reads a timed
+   * transcript (plus optional silence ranges) and the media sources, and plans
+   * the edit. The finished job's `output_data` holds the plan: an `Edl`
+   * (`"tighten"`), an {@link EdlClipSet} (`"clips"`), or a {@link ChapterSet}
+   * (`"chapters"`) — normalize it with {@link unwrapEditPlanOutput}.
+   *
+   * On a self-hosted install the request relays to nodaro.ai and needs the
+   * install connected (a 503 `code: "nodaro_connection_required"` otherwise);
+   * on nodaro.ai it runs directly.
+   */
+  editPlan(input) {
+    return this.client.request("POST", "/v1/edit-plan", {
+      body: {
+        mode: input.mode,
+        planTier: input.planTier,
+        transcript: input.transcript,
+        sources: input.sources,
+        ...input.silence !== void 0 ? { silence: input.silence } : {},
+        ...input.instructions !== void 0 ? { instructions: input.instructions } : {},
+        ...input.styleGuide !== void 0 ? { styleGuide: input.styleGuide } : {},
+        ...input.count !== void 0 ? { count: input.count } : {},
+        ...input.targetDurationSec !== void 0 ? { targetDurationSec: input.targetDurationSec } : {},
+        ...input.targetAspect !== void 0 ? { targetAspect: input.targetAspect } : {},
+        ...input.platform !== void 0 ? { platform: input.platform } : {},
+        ...input.workflowId !== void 0 ? { workflowId: input.workflowId } : {}
+      }
+    });
+  }
+  /**
+   * Remap a transcript through an EDL — a PURE client-side transform (NO
+   * request). Returns a new transcript whose word (and segment) timings are on
+   * the EDL's rendered output clock, dropping words that fall in cut regions and
+   * clipping straddlers. Runs `@nodaro/shared`'s canonical
+   * `remapTranscriptThroughEdl`, the same remap `applyEdl` performs server-side.
+   */
+  remapTranscript(edl, transcript) {
+    return shared.remapTranscriptThroughEdl(edl, transcript);
+  }
+};
+var SDK_VERSION = "2.16.0" ;
 var CLIENT_HEADER = "X-Nodaro-Client";
 var isBrowser = () => typeof window !== "undefined" && typeof window.document !== "undefined";
 var NodaroClient = class _NodaroClient {
@@ -3509,6 +3701,7 @@ var NodaroClient = class _NodaroClient {
   tutorials;
   organizations;
   workspaces;
+  edit;
   /** The workspace this client acts in; undefined = the personal space. */
   workspaceId;
   constructor(opts) {
@@ -3556,6 +3749,7 @@ var NodaroClient = class _NodaroClient {
     this.tutorials = new TutorialsResource(this);
     this.organizations = new OrganizationsResource(this);
     this.workspaces = new WorkspacesResource(this);
+    this.edit = new EditResource(this);
   }
   /**
    * A client that acts in `workspaceId`, sharing this one's auth and config.
@@ -3788,6 +3982,10 @@ Object.defineProperty(exports, "nodeStateMayCarryOutput", {
   enumerable: true,
   get: function () { return shared.nodeStateMayCarryOutput; }
 });
+Object.defineProperty(exports, "unwrapEditPlanOutput", {
+  enumerable: true,
+  get: function () { return shared.unwrapEditPlanOutput; }
+});
 Object.defineProperty(exports, "PEOPLE", {
   enumerable: true,
   get: function () { return prompts.PEOPLE; }
@@ -3815,6 +4013,7 @@ exports.CopilotResource = CopilotResource;
 exports.CreaturesResource = CreaturesResource;
 exports.CreditsResource = CreditsResource;
 exports.DeveloperAppsResource = DeveloperAppsResource;
+exports.EditResource = EditResource;
 exports.ExecutionsResource = ExecutionsResource;
 exports.ForbiddenError = ForbiddenError;
 exports.InsufficientCreditsError = InsufficientCreditsError;