@koda-sl/baker-cli 0.91.0 → 0.92.1

package/dist/cli.js

@@ -9,7 +9,7 @@ import {
   defaultRegistry,
   generateCatalog,
   validateCanvasDeep
-} from "./chunk-LMVDA3EZ.js";
+} from "./chunk-RCPMJKI7.js";
 
 // src/cli.ts
 import { defineCommand as defineCommand141, runMain } from "citty";
@@ -8369,6 +8369,18 @@ async function detectSceneCutsPySceneDetect(filePath, opts = {}) {
 }
 
 // src/engine/scaffold/video.ts
+import { toCardinal as nwAr } from "n2words/ar-SA";
+import { toCardinal as nwDe } from "n2words/de-DE";
+import { toCardinal as nwEn } from "n2words/en-US";
+import { toCardinal as nwEs } from "n2words/es-ES";
+import { toCardinal as nwFr } from "n2words/fr-FR";
+import { toCardinal as nwHi } from "n2words/hi-IN";
+import { toCardinal as nwIt } from "n2words/it-IT";
+import { toCardinal as nwJa } from "n2words/ja-JP";
+import { toCardinal as nwKo } from "n2words/ko-KR";
+import { toCardinal as nwNl } from "n2words/nl-NL";
+import { toCardinal as nwPl } from "n2words/pl-PL";
+import { toCardinal as nwPt } from "n2words/pt-PT";
 import { z as z3 } from "zod";
 
 // src/engine/scaffold/lib/shoot-modes.ts
@@ -8480,6 +8492,14 @@ var XFADE_BY_TYPE = {
   swipe: "wipeleft",
   zoom: "zoomin"
 };
+var DEFAULT_VIDEO_RESOLUTION = "1080p";
+var VIDEO_MODELS_WITH_RESOLUTION = new Set(
+  Object.entries(MODEL_REGISTRY.video_generate).filter(([, spec]) => "resolution" in spec.params).map(([id]) => id)
+);
+function videoResolutionParam(videoModel, resolution) {
+  if (!VIDEO_MODELS_WITH_RESOLUTION.has(videoModel)) return {};
+  return { resolution: resolution ?? DEFAULT_VIDEO_RESOLUTION };
+}
 var WORDS_PER_SECOND = 2.5;
 function estSpeechS(text) {
   const words = text.trim().split(/\s+/).filter(Boolean).length;
@@ -8697,12 +8717,21 @@ var VideoBlueprint = z3.object({
       // reference track. We never reuse it — only style the regenerated bed.
       identified_track: z3.object({ title: z3.string().optional(), artist: z3.string().optional() }).loose().nullish()
     }).loose().optional(),
-    cast: z3.array(z3.object({ id: z3.string().optional(), description: z3.string().optional() }).loose()).optional(),
+    cast: z3.array(
+      z3.object({
+        id: z3.string().optional(),
+        description: z3.string().optional(),
+        // The deconstruct's note on the target-market localization (e.g. "native
+        // French speaker") — read to derive the spoken-track language code.
+        market_localization_note: z3.string().optional()
+      }).loose()
+    ).optional(),
     voiceover: z3.object({
       // on_camera | mixed → mouths are on screen (lip-sync candidates);
       // voiceover | none → narration over the picture (no lip-sync).
       mode: z3.string().optional(),
-      voice_description: z3.string().optional()
+      voice_description: z3.string().optional(),
+      persona: z3.string().optional()
     }).loose().optional()
   }).loose().optional(),
   scenes: z3.array(Scene).min(1)
@@ -8885,11 +8914,18 @@ function slotsForFrame(slots, sceneIndex, edge) {
   return slots.filter((s) => s.presence.get(sceneIndex)?.has(edge));
 }
 var ACTOR_SHEET_MODEL = "google/gemini-3-pro-image-preview";
-function applyActorSheets(slots, nodes) {
+var SHEET_SUBJECT_TYPE = {
+  person: "person",
+  animal: "character",
+  product: "product",
+  location: "location"
+};
+function buildElementSheets(slots, nodes) {
   for (const slot of slots) {
-    const t = slot.type.toLowerCase();
-    if (t !== "person" && t !== "animal") continue;
-    if (slot.presence.size < 2) continue;
+    const subjectType = SHEET_SUBJECT_TYPE[slot.type.toLowerCase()];
+    if (!subjectType) continue;
+    if (slot.sameAs) continue;
+    if (slot.presence.size < 1) continue;
     const sheetId = `${slot.id}_sheet`;
     nodes.push({
       id: sheetId,
@@ -8899,11 +8935,15 @@ function applyActorSheets(slots, nodes) {
       params: {
         model: ACTOR_SHEET_MODEL,
         subject_description: slot.description ?? `the ${slot.type}`,
-        subject_type: t === "person" ? "person" : "character",
-        image_size: "2K"
+        subject_type: subjectType,
+        // 4K: the sheet packs up to 8 cells (angles + tight face/detail close-ups), and
+        // it's the ONE reference every frame grounds on — per-cell sharpness here
+        // propagates to every clip, so it's worth the highest tier on this single asset.
+        image_size: "4K"
       }
     });
     slot.ref = `$ref:${sheetId}.sheet`;
+    slot.sheetBacked = true;
   }
 }
 function slotsForScene(slots, sceneIndex) {
@@ -8914,7 +8954,7 @@ function buildFramePrompt(edge, sceneIndex, framePrompt, present, hasAnchor, mod
   const legend = [
     ...present.map((s) => `- ${s.label} \u2014 ${roleForSlot(s)}`),
     ...hasAnchor ? [
-      "- ORIGINAL_FRAME \u2014 use ONLY for composition, framing, pose, and proportions. IGNORE its text, its logo, its brand name, and its colors entirely \u2014 it is a DIFFERENT brand's footage, here only to anchor layout/pose, never identity or palette."
+      "- ORIGINAL_FRAME \u2014 use ONLY for composition: framing, camera angle, shot size, subject placement, pose, and proportions. IGNORE its text, logo, brand name, colors, AND the identity of every person/animal/object in it \u2014 those come from the labeled reference images above, never from this frame. It is a DIFFERENT brand's footage with DIFFERENT actors, here ONLY to anchor where things sit and how the shot is framed (e.g. a profile/side angle stays a profile/side angle), never who they are or what palette to use."
     ] : []
   ].join("\n");
   const description = framePrompt?.trim() || `the ${edge} frame of scene ${sceneIndex + 1} \u2014 describe the full composition, subjects, setting, action, lighting, and palette here. (Edit this line to change ONLY this frame.)`;
@@ -9003,11 +9043,12 @@ function ingestFrameRef(url, edge, ctx, nodes) {
 function buildFrameRef(edge, url, framePrompt, present, ctx, nodes) {
   const tag = ctx.tag ?? "";
   if (ctx.reuse && url) return ingestFrameRef(url, edge, ctx, nodes);
-  const hasPersonOrAnimal = present.some((s) => {
+  const castSlots = present.filter((s) => {
     const t = s.type.toLowerCase();
     return t === "person" || t === "animal";
   });
-  const useOriginalAnchor = Boolean(url) && !hasPersonOrAnimal;
+  const castIdentityLocked = castSlots.every((s) => s.sheetBacked);
+  const useOriginalAnchor = Boolean(url) && (castSlots.length === 0 || castIdentityLocked);
   const hasOriginal = useOriginalAnchor;
   const originalRef = useOriginalAnchor && url ? ingestFrameRef(url, edge, ctx, nodes) : void 0;
   const reference = [...present.map((s) => s.ref), ...originalRef ? [originalRef] : []];
@@ -9039,17 +9080,18 @@ function seedanceAudioLine(scene, mode, audio, nativeLine) {
   }
   return null;
 }
-function buildSeedancePrompt(scene, sceneIndex, present, mode, audio, nativeLine) {
+function buildSeedancePrompt(scene, sceneIndex, present, mode, audio, nativeLine, nativeLang) {
+  const loc = (s) => nativeLine ? localizeNumeralsForNative(s, nativeLang) : s;
   const parts = [];
   const summary = scene.summary?.trim();
-  parts.push(summary ? `Scene ${sceneIndex + 1}: ${summary}` : `Scene ${sceneIndex + 1}`);
-  if (scene.action_detail) parts.push(`Action: ${scene.action_detail}`);
+  parts.push(summary ? `Scene ${sceneIndex + 1}: ${loc(summary)}` : `Scene ${sceneIndex + 1}`);
+  if (scene.action_detail) parts.push(`Action: ${loc(scene.action_detail)}`);
   const cm = scene.camera_motion;
   if (cm) {
     const camera = [cm.movement, cm.detail].filter(Boolean).join(" \u2014 ");
     if (camera) parts.push(`Camera: ${camera}`);
   }
-  if (scene.motion_prompt) parts.push(`Motion: ${scene.motion_prompt}`);
+  if (scene.motion_prompt) parts.push(`Motion: ${loc(scene.motion_prompt)}`);
   if (present.length > 0) {
     parts.push(
       `Keep these consistent with their references: ${present.map((s) => `${s.label} (${s.description ?? s.type})`).join("; ")}`
@@ -9057,7 +9099,7 @@ function buildSeedancePrompt(scene, sceneIndex, present, mode, audio, nativeLine
   }
   if (nativeLine) {
     parts.push(
-      `The person speaks to camera. Lip-sync follows the dialogue verbatim; put delivery/emotion cues in [brackets]. Dialogue: "${nativeLine}"`
+      `The person speaks to camera. Lip-sync follows the dialogue verbatim; put delivery/emotion cues in [brackets]. Dialogue: "${loc(nativeLine)}"`
     );
   } else {
     const lines = (scene.dialogue ?? []).map((d) => d.line?.trim()).filter((l) => Boolean(l));
@@ -9065,7 +9107,7 @@ function buildSeedancePrompt(scene, sceneIndex, present, mode, audio, nativeLine
       parts.push(`Spoken context (do not render as audio): ${lines.map((l) => `"${l}"`).join(" ")}`);
   }
   const transcript = (scene.transcript_slice ?? []).map((w) => w.text?.trim()).filter(Boolean).join(" ").trim();
-  if (transcript) parts.push(`Transcript: ${transcript}`);
+  if (transcript) parts.push(`Transcript: ${loc(transcript)}`);
   const audioLine = seedanceAudioLine(scene, mode, audio, nativeLine);
   if (audioLine) parts.push(audioLine);
   parts.push(
@@ -9176,8 +9218,17 @@ function buildPerSpeakerVoiceConversion(segments, totalMs, nodes) {
 function emitSceneClip(i, scene, present, mode, nativeTurn, ambientBroll, frames, lengths, out, opts, nodes, tag = "") {
   const clipParams = {
     model: opts.videoModel,
-    prompt: buildSeedancePrompt(scene, i, present, mode, Boolean(nativeTurn) || ambientBroll, nativeTurn?.text),
+    prompt: buildSeedancePrompt(
+      scene,
+      i,
+      present,
+      mode,
+      Boolean(nativeTurn) || ambientBroll,
+      nativeTurn?.text,
+      opts.nativeLang
+    ),
     duration: lengths.genDur,
+    ...videoResolutionParam(opts.videoModel, opts.resolution),
     // Native talking scene → Seedance generates the spoken audio + lip-sync; an opt-in
     // ambient b-roll beat generates diegetic ambient only; otherwise the clip is silent.
     generate_audio: Boolean(nativeTurn) || ambientBroll
@@ -9281,7 +9332,7 @@ function buildCompositeScene(layout, regions, comp, scene, i, present, mode, nat
       { first, last },
       lengths,
       null,
-      { ar: opts.ar, videoModel: opts.videoModel },
+      { ar: opts.ar, videoModel: opts.videoModel, resolution: opts.resolution, nativeLang: opts.nativeLang },
       nodes,
       tag
     );
@@ -9413,6 +9464,65 @@ var LANGUAGE_WORDS = [
   [/\b(hindi)\b/, "hindi"],
   [/\b(polish)\b/, "polish"]
 ];
+var LANGUAGE_ISO = {
+  french: "fr",
+  spanish: "es",
+  english: "en",
+  german: "de",
+  italian: "it",
+  portuguese: "pt",
+  dutch: "nl",
+  arabic: "ar",
+  japanese: "ja",
+  korean: "ko",
+  hindi: "hi",
+  polish: "pl"
+};
+function languageHaystacks(blueprint) {
+  const vo = blueprint.global?.voiceover;
+  const cast = blueprint.global?.cast ?? [];
+  const dialogue = blueprint.scenes.flatMap((s) => s.dialogue ?? []);
+  return [
+    vo?.voice_description,
+    vo?.persona,
+    ...cast.flatMap((c) => [c.market_localization_note, c.description]),
+    ...dialogue.map((l) => l.voice_description)
+  ].filter((s) => Boolean(s));
+}
+function deriveTtsLanguageCode(blueprint) {
+  for (const text of languageHaystacks(blueprint)) {
+    const name = parseVoiceTraits(text).language;
+    if (name && LANGUAGE_ISO[name]) return LANGUAGE_ISO[name];
+  }
+  return void 0;
+}
+var INTEGER_SPELLERS = {
+  fr: nwFr,
+  es: nwEs,
+  en: nwEn,
+  de: nwDe,
+  it: nwIt,
+  pt: nwPt,
+  nl: nwNl,
+  pl: nwPl,
+  ar: nwAr,
+  ja: nwJa,
+  ko: nwKo,
+  hi: nwHi
+};
+function spellNumber(langCode, n) {
+  const spell = langCode ? INTEGER_SPELLERS[langCode] : void 0;
+  if (!spell || !Number.isFinite(n)) return String(n);
+  try {
+    return spell(n);
+  } catch {
+    return String(n);
+  }
+}
+function localizeNumeralsForNative(text, langCode) {
+  if (!langCode || !INTEGER_SPELLERS[langCode]) return text;
+  return text.replace(/(?<![\w.,-])\d{1,9}(?![\w.,-])/g, (m) => spellNumber(langCode, Number.parseInt(m, 10)));
+}
 function parseVoiceTraits(description) {
   const d = description.toLowerCase();
   const out = {};
@@ -9431,14 +9541,14 @@ function isOnCameraSpeaker(speaker, casts, cameraOn) {
   if (NARRATOR_SPEAKERS.has(speaker.toLowerCase())) return false;
   return casts.has(speaker);
 }
-function makePresenterPresent(slots, canonical) {
+function makePresenterPresent(slots, canonical, opts = {}) {
   const personSlots = slots.filter((s) => s.type.toLowerCase() === "person");
   const bySpeaker = /* @__PURE__ */ new Map();
   for (const slot of personSlots) if (slot.castId) bySpeaker.set(canonical(slot.castId), slot.presence);
-  const solePerson = personSlots.length === 1 ? personSlots[0].presence : null;
+  const solePerson = !opts.strict && personSlots.length === 1 ? personSlots[0].presence : null;
   return (speaker, sceneIndex) => {
     const presence = bySpeaker.get(speaker) ?? solePerson;
-    if (!presence) return true;
+    if (!presence) return opts.strict ? false : true;
     return presence.has(sceneIndex);
   };
 }
@@ -9457,16 +9567,18 @@ function collapseVoiceover(blueprint) {
   const presenter = [...presenters][0];
   return (speaker) => NARRATOR_SPEAKERS.has(speaker.toLowerCase()) ? presenter : speaker;
 }
-function buildPhrases(blueprint, canonical, compositeScenes, presenterPresent) {
+function buildPhrases(blueprint, canonical, compositeScenes, presenterPresent, presentStrict) {
   const casts = castIdSet(blueprint);
   const cameraOn = onCameraDialogue(blueprint);
   const sceneEndS = (i) => blueprint.scenes[i]?.end_s ?? blueprint.scenes[i]?.start_s ?? 0;
   const multiSpeaker = /* @__PURE__ */ new Set();
   blueprint.scenes.forEach((scene, i) => {
-    const onCam = new Set(
+    const onCamAll = new Set(
       (scene.dialogue ?? []).map((l) => l.speaker ?? "voiceover").filter((sp) => isOnCameraSpeaker(sp, casts, cameraOn))
     );
-    if (onCam.size >= 2) multiSpeaker.add(i);
+    const onCamPresent = [...onCamAll].filter((sp) => presentStrict(canonical(sp), i));
+    const effective = onCamPresent.length > 0 ? new Set(onCamPresent) : onCamAll;
+    if (effective.size >= 2) multiSpeaker.add(i);
   });
   const lines = blueprint.scenes.flatMap(
     (scene, sceneIndex) => compositeScenes.has(sceneIndex) ? [] : (scene.dialogue ?? []).filter((l) => Boolean(l.line?.trim())).map((l) => {
@@ -9605,8 +9717,9 @@ function emitPhraseClip(phrase, voiceNode, env, nodes, out) {
   const genDur = ceilToSeedance(phraseLen);
   const clipParams = {
     model: env.opts.videoModel,
-    prompt: buildSeedancePrompt(anchorScene, anchor, present, mode, true, phrase.text),
+    prompt: buildSeedancePrompt(anchorScene, anchor, present, mode, true, phrase.text, env.ttsLanguageCode),
     duration: genDur,
+    ...videoResolutionParam(env.opts.videoModel, env.opts.resolution),
     generate_audio: true
   };
   if (env.ar) clipParams.aspect_ratio = env.ar;
@@ -9666,7 +9779,7 @@ function emitPhraseClip(phrase, voiceNode, env, nodes, out) {
     });
   }
 }
-function emitPhraseTts(phrase, voiceNode, idx, used, nodes, out) {
+function emitPhraseTts(phrase, voiceNode, idx, used, nodes, out, languageCode) {
   let id = sanitizeId2(`vo_ph${idx}_${phrase.speaker}`, `vo_ph${idx}`);
   while (used.has(id)) id = `${id}_x`;
   used.add(id);
@@ -9674,7 +9787,12 @@ function emitPhraseTts(phrase, voiceNode, idx, used, nodes, out) {
     id,
     type: "tts",
     inputs: { voice_ref: `$ref:${voiceNode}.voice_id` },
-    params: { model: FIXED_TTS_MODEL, text: phrase.text, voice: "{{voice_ref}}" }
+    params: {
+      model: FIXED_TTS_MODEL,
+      text: phrase.text,
+      voice: "{{voice_ref}}",
+      ...languageCode ? { language_code: languageCode } : {}
+    }
   });
   out.voTracks.push({ slot: id, ref: `$ref:${id}.audio`, start_s: phrase.start_s, end_s: phrase.end_s, kind: "vo" });
   out.voSegments.push({
@@ -9717,17 +9835,34 @@ function emitCompositeInTimeline(composite, scene, i, isLast, env, canonical, en
     nativeTurn,
     lengths,
     lengths.out,
-    { ar: env.ar, reuse: env.reuse, imageModel: env.opts.imageModel, videoModel: env.opts.videoModel },
+    {
+      ar: env.ar,
+      reuse: env.reuse,
+      imageModel: env.opts.imageModel,
+      videoModel: env.opts.videoModel,
+      resolution: env.opts.resolution,
+      nativeLang: env.ttsLanguageCode
+    },
     nodes,
     out.voTracks,
     out.nativeSegments,
     out.clips
   );
   if (!nativeTurn && distinctSpeakers.size >= 2) {
-    emitCompositeMultiSpeakerVoice(onCam, scene, i, canonical, ensureVoiceNode, usedVoIds, nodes, out);
+    emitCompositeMultiSpeakerVoice(
+      onCam,
+      scene,
+      i,
+      canonical,
+      ensureVoiceNode,
+      usedVoIds,
+      nodes,
+      out,
+      env.ttsLanguageCode
+    );
   }
 }
-function emitCompositeMultiSpeakerVoice(onCam, scene, i, canonical, ensureVoiceNode, usedVoIds, nodes, out) {
+function emitCompositeMultiSpeakerVoice(onCam, scene, i, canonical, ensureVoiceNode, usedVoIds, nodes, out, languageCode) {
   const bySpeaker = /* @__PURE__ */ new Map();
   for (const l of onCam) {
     const speaker = canonical(l.speaker ?? "voiceover");
@@ -9759,7 +9894,8 @@ function emitCompositeMultiSpeakerVoice(onCam, scene, i, canonical, ensureVoiceN
       i,
       usedVoIds,
       nodes,
-      out
+      out,
+      languageCode
     );
   }
 }
@@ -9806,7 +9942,7 @@ function emitBrollScene(scene, i, isLast, env, nodes, out, prevEndFrame) {
     { first, last },
     { dur: lengths.dur, trimTarget: lengths.trimTarget, genDur: lengths.genDur },
     lengths.out,
-    { ar: env.ar, videoModel: env.opts.videoModel },
+    { ar: env.ar, videoModel: env.opts.videoModel, resolution: env.opts.resolution, nativeLang: env.ttsLanguageCode },
     nodes
   );
   if (ambientBroll) {
@@ -9842,7 +9978,8 @@ function buildTimeline(blueprint, slots, opts, nodes) {
     reuse,
     cameraOn: onCameraDialogue(blueprint),
     casts: castIdSet(blueprint),
-    ingestCache: /* @__PURE__ */ new Map()
+    ingestCache: /* @__PURE__ */ new Map(),
+    ttsLanguageCode: deriveTtsLanguageCode(blueprint)
   };
   const out = {
     clips: [],
@@ -9853,7 +9990,8 @@ function buildTimeline(blueprint, slots, opts, nodes) {
     sceneSlice: /* @__PURE__ */ new Map()
   };
   const presenterPresent = makePresenterPresent(slots, canonical);
-  const phrases = buildPhrases(blueprint, canonical, compositeScenes, presenterPresent);
+  const presentStrict = makePresenterPresent(slots, canonical, { strict: true });
+  const phrases = buildPhrases(blueprint, canonical, compositeScenes, presenterPresent, presentStrict);
   const usedVoIds = /* @__PURE__ */ new Set();
   const claimed = /* @__PURE__ */ new Set();
   phrases.forEach((phrase, k) => {
@@ -9863,7 +10001,7 @@ function buildTimeline(blueprint, slots, opts, nodes) {
       for (const s of available) claimed.add(s);
       emitPhraseClip({ ...phrase, shownScenes: available }, voiceNode, env, nodes, out);
     } else {
-      emitPhraseTts(phrase, voiceNode, k, usedVoIds, nodes, out);
+      emitPhraseTts(phrase, voiceNode, k, usedVoIds, nodes, out, env.ttsLanguageCode);
     }
   });
   const lastIndex = blueprint.scenes.length - 1;
@@ -10200,7 +10338,7 @@ function scaffoldVideoCanvas(input, elementsInput, opts) {
       params: { source: "path", path: todoPath2(elements[i], slot.label), expect: "image" }
     });
   });
-  applyActorSheets(slots, nodes);
+  buildElementSheets(slots, nodes);
   const { clips, voTracks, vo_segments, talking_scenes } = buildTimeline(blueprint, slots, opts, nodes);
   let videoRef = buildSpine(clips, nodes);
   let videoNode = "spine";
@@ -10309,9 +10447,27 @@ function buildVideoMeta(blueprint, meta) {
     duration_s: blueprint.source?.duration_s ?? lastSceneEnd(blueprint),
     vo_segments: [...meta.vo_segments].sort((a, b) => a.start_s - b.start_s),
     talking_scenes: meta.talking_scenes,
+    lip_sync_caution: buildLipSyncCaution(meta.vo_segments),
     motion_board: buildMotionBoard(blueprint)
   };
 }
+function buildLipSyncCaution(segments) {
+  const out = [];
+  const byScene = /* @__PURE__ */ new Map();
+  for (const s of segments) {
+    const arr = byScene.get(s.scene) ?? [];
+    arr.push(s);
+    byScene.set(s.scene, arr);
+  }
+  for (const [scene, segs] of [...byScene.entries()].sort((a, b) => a[0] - b[0])) {
+    const nativeSpeakers = new Set(segs.filter((s) => s.slot.endsWith("_conv")).map((s) => s.speaker));
+    for (const speaker of nativeSpeakers) {
+      const ttsOver = segs.filter((s) => !s.slot.endsWith("_conv") && s.speaker === speaker).map((s) => s.slot);
+      if (ttsOver.length > 0) out.push({ scene, speaker, tts_over_native: ttsOver });
+    }
+  }
+  return out;
+}
 function sceneSpokenText(scene) {
   return (scene.dialogue ?? []).map((d) => d.line?.trim()).filter((l) => Boolean(l)).join(" ") || null;
 }
@@ -10743,7 +10899,11 @@ var scaffoldVideoCommand = defineCommand76({
     "deconstruct-model": { type: "string", description: "Override the video_deconstruct model id" },
     "select-model": { type: "string", description: "Override the text_generate model id for element selection" },
     "image-model": { type: "string", description: "Override the image_generate model id for frames" },
-    "video-model": { type: "string", description: "Override the video_generate model id for clips" }
+    "video-model": { type: "string", description: "Override the video_generate model id for clips" },
+    resolution: {
+      type: "string",
+      description: `Output resolution for generated clips (e.g. "1080p"). Default 1080p \u2014 the highest the video model supports \u2014 so clips keep the keyframe sharpness instead of the model's low default.`
+    }
   },
   async run({ args }) {
     const videoPath = path5.resolve(String(args.file));
@@ -10795,7 +10955,8 @@ var scaffoldVideoCommand = defineCommand76({
       transcriptPath: captions.transcriptPath,
       blueprintPath,
       frames,
-      ambient: Boolean(args.ambient)
+      ambient: Boolean(args.ambient),
+      ...args.resolution ? { resolution: String(args.resolution) } : {}
     };
     let canvas;
     let report;