@koda-sl/baker-cli 0.281.9-dev.1f1c09c80 → 0.281.11-dev.0c0293389

package/dist/cli.js

@@ -58,7 +58,7 @@ import {
   ulid,
   validateCanvasDeep,
   ytDlpBlockSignal
-} from "./chunk-XSGYVQXF.js";
+} from "./chunk-YNUBWTLR.js";
 import {
   csvOrJson,
   daysAgoIso,
@@ -27372,6 +27372,11 @@ var VideoBlueprint = z27.object({
       // on_camera | mixed → mouths are on screen (lip-sync candidates);
       // voiceover | none → narration over the picture (no lip-sync).
       mode: z27.string().optional(),
+      // Which variety of the language the speech is in. Stated because the model
+      // picks one otherwise: ads set in Spain came back in Latin American Spanish,
+      // and the pinned peninsular voice could not undo it — the re-voicing is
+      // speech-to-speech, so it carries the accent across and swaps only the timbre.
+      accent: z27.string().optional(),
       voice_description: z27.string().optional(),
       persona: z27.string().optional()
     }).loose().optional(),
@@ -27953,10 +27958,10 @@ function visualBriefParts(scene, sceneIndex, plate) {
   }
   return parts;
 }
-function spokenTextParts(scene, nativeLine, loc, profile) {
+function spokenTextParts(scene, nativeLine, loc, profile, spokenAccent) {
   const parts = [];
   if (nativeLine) {
-    parts.push(profile.dialogueDirective(loc(nativeLine)));
+    parts.push(profile.dialogueDirective(loc(nativeLine), spokenAccent));
   } else {
     const lines = (scene.dialogue ?? []).map((d) => d.line?.trim()).filter((l) => Boolean(l));
     if (lines.length > 0)
@@ -27966,7 +27971,7 @@ function spokenTextParts(scene, nativeLine, loc, profile) {
   if (transcript) parts.push(`Transcript: ${loc(transcript)}`);
   return parts;
 }
-function buildSeedancePrompt(scene, sceneIndex, present2, mode, audio, nativeLine, nativeLang, uiRouted, videoModel) {
+function buildSeedancePrompt(scene, sceneIndex, present2, mode, audio, nativeLine, nativeLang, uiRouted, videoModel, spokenAccent) {
   const profile = (videoModel ? clipProfileFor(videoModel) : void 0) ?? SEEDANCE_PROFILE;
   const loc = (s) => nativeLine ? localizeNumeralsForNative(s, nativeLang) : s;
   const routed = uiRouted ?? (compositeRegionsOf(scene) !== null && isUiOnlyComposite(compositeRegionsOf(scene) ?? []));
@@ -27978,7 +27983,7 @@ function buildSeedancePrompt(scene, sceneIndex, present2, mode, audio, nativeLin
       `Keep these consistent with their references: ${refs.map((s) => `${s.label} (${s.description ?? s.type})`).join("; ")}`
     );
   }
-  parts.push(...spokenTextParts(scene, nativeLine, loc, profile));
+  parts.push(...spokenTextParts(scene, nativeLine, loc, profile, spokenAccent));
   const audioLine = seedanceAudioLine(scene, mode, audio, nativeLine);
   if (audioLine) parts.push(audioLine);
   parts.push(
@@ -28123,7 +28128,8 @@ function emitSceneClip(i, scene, present2, mode, nativeTurn, ambientBroll, frame
       nativeTurn?.text,
       opts.nativeLang,
       opts.uiRouted,
-      opts.videoModel
+      opts.videoModel,
+      opts.spokenAccent
     ),
     ...videoDurationParam(opts.videoModel, lengths.genDur),
     ...videoResolutionParam(opts.videoModel, opts.resolution),
@@ -28310,7 +28316,8 @@ function buildCompositeScene(layout, regions, comp, scene, i, present2, mode, na
         genAr: opts.genAr,
         videoModel: opts.videoModel,
         resolution: opts.resolution,
-        nativeLang: opts.nativeLang
+        nativeLang: opts.nativeLang,
+        spokenAccent: opts.spokenAccent
       },
       nodes,
       tag
@@ -28889,7 +28896,8 @@ function emitPhraseClip(phrase, voiceNode, env, nodes, out) {
       phrase.text,
       env.ttsLanguageCode,
       env.uiRouted.has(anchor),
-      env.opts.videoModel
+      env.opts.videoModel,
+      env.spokenAccent
     ),
     ...videoDurationParam(env.opts.videoModel, genDur),
     ...videoResolutionParam(env.opts.videoModel, env.opts.resolution),
@@ -29039,7 +29047,8 @@ function emitCompositeInTimeline(composite, scene, i, isLast, env, canonical2, e
       imageModel: env.opts.imageModel,
       videoModel: env.opts.videoModel,
       resolution: env.opts.resolution,
-      nativeLang: env.ttsLanguageCode
+      nativeLang: env.ttsLanguageCode,
+      spokenAccent: env.spokenAccent
     },
     env.clock,
     nodes,
@@ -29175,6 +29184,7 @@ function emitBrollScene(scene, i, isLast, env, nodes, out, prevEndFrame) {
       videoModel: env.opts.videoModel,
       resolution: env.opts.resolution,
       nativeLang: env.ttsLanguageCode,
+      spokenAccent: env.spokenAccent,
       uiRouted: env.uiRouted.has(i)
     },
     nodes,
@@ -29236,7 +29246,8 @@ function buildTimeline(blueprint, slots, opts, nodes) {
     uiRouted,
     graphicScenes: fullFrameGraphicScenes(blueprint),
     surfaceIngests: /* @__PURE__ */ new Map(),
-    ttsLanguageCode: deriveTtsLanguageCode(blueprint)
+    ttsLanguageCode: deriveTtsLanguageCode(blueprint),
+    spokenAccent: blueprint.global?.voiceover?.accent?.trim() || void 0
   };
   const out = {
     routes: /* @__PURE__ */ new Map(),
@@ -33033,6 +33044,11 @@ function staticTranscript(spec) {
   }
   return out;
 }
+function spokenVariety(spec) {
+  const market = marketFor(spec);
+  if (!market) return null;
+  return `the local accent of ${market} \u2014 the way people there actually speak, not a neighbouring country's variety of the same language`;
+}
 function musicBed(spec) {
   if (spec.music === false) return null;
   if (typeof spec.music === "string") return spec.music;
@@ -33121,6 +33137,7 @@ function adSpecToBlueprint(spec) {
       transcript: spec.beats.map((b) => b.say).join(" "),
       voiceover: {
         mode: voiceoverMode(spec, nativeSpeech),
+        ...spokenVariety(spec) ? { accent: spokenVariety(spec) } : {},
         ...spec.voice?.description ? { voice_description: spec.voice.description } : {}
       },
       ...spec.cast ? {
@@ -33326,11 +33343,29 @@ function svgIsLiveText(svg) {
   return /<text[\s>]/i.test(svg);
 }
 async function markIsReal(markPath, root = ".") {
-  const brandLogos = path22.join(root, BRAND_DIR, "logos");
-  const brandFiles = await readdir8(brandLogos).catch(() => null);
-  if (brandFiles) return await markDirectoryIsReal(brandLogos);
+  const brandLogos = await findBrandLogosDir(markPath, root);
+  if (brandLogos) return await markDirectoryIsReal(brandLogos);
   return await markDirectoryIsReal(path22.dirname(markPath));
 }
+async function findBrandLogosDir(markPath, root) {
+  const candidates = [];
+  let dir = path22.resolve(path22.dirname(markPath));
+  for (; ; ) {
+    candidates.push(path22.join(dir, BRAND_DIR, "logos"));
+    const parent = path22.dirname(dir);
+    if (parent === dir) break;
+    dir = parent;
+  }
+  candidates.push(path22.join(root, BRAND_DIR, "logos"));
+  for (const candidate of candidates) {
+    if (await readdir8(candidate).then(
+      () => true,
+      () => false
+    ))
+      return candidate;
+  }
+  return null;
+}
 async function markDirectoryIsReal(dir) {
   const files = await readdir8(dir).catch(() => null);
   if (!files) return false;