@koda-sl/baker-cli 0.275.0-dev.1f1c09c80 → 0.277.0-dev.1f1c09c80

package/dist/cli.js

@@ -58,7 +58,7 @@ import {
   ulid,
   validateCanvasDeep,
   ytDlpBlockSignal
-} from "./chunk-4IJX4R4J.js";
+} from "./chunk-AUOSDGWY.js";
 import {
   csvOrJson,
   daysAgoIso,
@@ -3699,14 +3699,22 @@ var avatarSummarySchema = z7.object({
   subjectDescription: z7.string(),
   /** The identity sheet to wire into `--reference`. Absent until the build settles. */
   sheetUrl: z7.string().optional(),
+  /**
+   * The avatar's pinned voice.
+   *
+   * A casting essential, not a profile detail: an avatar is a face AND a voice, and
+   * casting the face without it is what makes a presenter speak in somebody else's.
+   * It lived only behind `--full`, so an ad that grounded every frame on the identity
+   * sheet still had its lines read by a voice cast from a prose description.
+   */
+  voiceId: z7.string().optional(),
+  voiceDescription: z7.string().optional(),
   coverUrl: z7.string().optional(),
   errorMessage: z7.string().optional()
 });
 var avatarDetailSchema = avatarSummarySchema.extend({
   profile: avatarProfileSchema,
   sourceUrls: z7.array(z7.string()),
-  voiceId: z7.string().optional(),
-  voiceDescription: z7.string().optional(),
   createdAt: z7.number()
 });
 var avatarsListRequestSchema = z7.object({
@@ -28705,7 +28713,7 @@ function buildPhrases(blueprint, canonical2, compositeScenes, presenterPresent,
   flush();
   return phrases;
 }
-function makeVoiceFactory(blueprint, canonical2, nodes, voiceLanguage) {
+function makeVoiceFactory(blueprint, canonical2, nodes, voiceLanguage, pinnedVoiceId) {
   const bySpeaker = /* @__PURE__ */ new Map();
   const describe = (speaker) => {
     for (const scene of blueprint.scenes)
@@ -28721,7 +28729,9 @@ function makeVoiceFactory(blueprint, canonical2, nodes, voiceLanguage) {
     const description = describe(speaker);
     const traits = parseVoiceTraits(description);
     if (voiceLanguage) traits.language = voiceLanguage;
-    nodes.push({ id, type: "voice_select", params: { description, ...traits } });
+    nodes.push(
+      pinnedVoiceId ? { id, type: "voice_select", params: { description, voice_id: pinnedVoiceId, ...traits } } : { id, type: "voice_select", params: { description, ...traits } }
+    );
     bySpeaker.set(speaker, id);
     return id;
   };
@@ -29122,7 +29132,7 @@ function buildTimeline(blueprint, slots, opts, nodes) {
     });
   }
   const canonical2 = collapseVoiceover(blueprint);
-  const ensureVoiceNode = makeVoiceFactory(blueprint, canonical2, nodes, opts.voiceLanguage);
+  const ensureVoiceNode = makeVoiceFactory(blueprint, canonical2, nodes, opts.voiceLanguage, opts.voiceId);
   const aspect = resolveAspect(blueprint.source?.aspect_ratio, opts.aspect, genAspectsFor(opts.videoModel));
   const env = {
     blueprint,
@@ -31002,7 +31012,7 @@ function classifyReelFindings(json) {
   if (!r.brand_mark_visible) {
     findings.push({
       severity: "warning",
-      what: "No brand mark is visible in any frame. `baker canvas scaffold-ad` draws one from `brand.logo` (the repo path in src/brand/BRAND.md) \u2014 top-left throughout and large on the closing card \u2014 so a mark missing here means the spec carried no logo, or the path did not resolve and the scaffold said so."
+      what: "No brand mark is visible in any frame. THIS IS NOT A RE-RENDER: the mark is drawn by the caption composition as an ordinary `<img>`, never generated as a picture \u2014 so fix the mark itself (a bad path, or an SVG whose geometry clips its own wordmark) or the `.brand-mark` / `.endcard-logo` rules in `<composition>/index.html`, then re-run. Every clip comes back from the cache and only the composition pass renders again, so it costs seconds and no generation."
     });
   }
   if (!r.shows_more_than_a_face) {
@@ -31429,7 +31439,7 @@ ${describeRewrites(healed.rewrites)}
       `Reviewed the finished video and found ${review.length} thing${review.length === 1 ? "" : "s"} to fix. Show the user what is wrong before you show them the ad.`,
       ...review.map((f) => `${f.severity === "blocking" ? "MUST FIX" : "SHOULD FIX"} \u2014 ${f.what}`),
       ...review.some((f) => f.scene) ? [
-        `Fix them, then re-run: edit the \`show\` of the scene files named above \u2014 a shot with a person handling equipment is where limbs and objects go wrong, so widen it, move the hands out of frame, or drop the person from that beat. Re-running re-renders ONLY the scenes you edited; everything else comes from the cache (this run: ${result.stats?.cached_nodes ?? 0} of ${result.stats?.total_nodes ?? 0} nodes were cached).`
+        `Anything about the MARK or the closing card is a composition fix, not a re-render: both are drawn as HTML by the caption composition, so editing it and re-running re-renders only that pass \u2014 the clips are cached and no generation is paid for again. Fix them, then re-run: edit the \`show\` of the scene files named above \u2014 a shot with a person handling equipment is where limbs and objects go wrong, so widen it, move the hands out of frame, or drop the person from that beat. Re-running re-renders ONLY the scenes you edited; everything else comes from the cache (this run: ${result.stats?.cached_nodes ?? 0} of ${result.stats?.total_nodes ?? 0} nodes were cached).`
       ] : []
     ] : [];
     if (poster) {
@@ -32775,7 +32785,8 @@ function adSpecCastElements(spec, avatar) {
 function motionPrompt(beat, isClosingCard, place, physics) {
   if (isClosingCard) return BRAND_PLATE;
   const shot = `${beat.show}${place}${physics}`;
-  return beat.on_camera ? shot : `${shot} Nobody in frame is speaking \u2014 mouths closed, no dialogue.`;
+  if (beat.on_camera) return shot;
+  return `${shot} Nobody in frame is speaking: mouths CLOSED and STILL throughout, no talking, no lip movement. Whoever is in shot is looking, reacting or working \u2014 never addressing the camera.`;
 }
 function staticTranscript(spec) {
   const out = [];
@@ -32864,6 +32875,7 @@ function adSpecToBlueprint(spec) {
 }
 
 // src/engine/scaffold/ad-brand-overlay.ts
+var markFallback = (name) => name.trim() ? ` onerror="this.replaceWith(Object.assign(document.createElement('span'),{className:this.className+' brand-wordmark',textContent:${JSON.stringify(name.trim())}}))"` : "";
 var escapeHtml2 = (s) => s.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;").replace(/"/g, "&quot;");
 function firstHex(palette) {
   const hex = (palette ?? []).find((h) => /^#?[0-9a-fA-F]{6}$/.test(h.trim()));
@@ -32883,12 +32895,14 @@ function buildAdBrandOverlay(opts) {
   ];
   if (logo) {
     parts.push(
-      `<img class="clip brand-mark" src="${escapeHtml2(logo.file)}" alt="${escapeHtml2(logo.alt)}" data-start="0" data-dur="${totalS}" />`
+      `<img class="clip brand-mark" src="${escapeHtml2(logo.file)}" alt="${escapeHtml2(logo.alt)}" data-start="0" data-dur="${totalS}"${markFallback(logo.alt)} />`
     );
   }
   const cardBits = [];
   if (logo) {
-    cardBits.push(`<img class="endcard-logo" src="${escapeHtml2(logo.file)}" alt="${escapeHtml2(logo.alt)}" />`);
+    cardBits.push(
+      `<img class="endcard-logo" src="${escapeHtml2(logo.file)}" alt="${escapeHtml2(logo.alt)}"${markFallback(logo.alt)} />`
+    );
   }
   if (cta?.trim()) {
     const style = accent ? ` style="background:${escapeHtml2(accent)}"` : "";
@@ -32959,6 +32973,11 @@ function adBrandOverlayCss() {
     "      .endcard-logo { width: 62%; height: auto; max-height: 300px; object-fit: contain;",
     "        padding: 46px 56px; border-radius: 40px; background: rgba(255,255,255,0.96);",
     "        box-shadow: 0 10px 40px rgba(0,0,0,0.28); }",
+    "      /* The name, when the mark itself could not be drawn. */",
+    "      .brand-wordmark { display: inline-block; font-family: 'Arial Black', 'Helvetica Neue', Arial, sans-serif;",
+    "        font-weight: 900; color: #14231b; letter-spacing: -0.01em; }",
+    "      .brand-mark.brand-wordmark { font-size: 34px; line-height: 1; height: auto; }",
+    "      .endcard-logo.brand-wordmark { font-size: 96px; line-height: 1; width: auto; }",
     "      /* Capped and wrapping, not sized to its content: measured in a real browser,",
     "         `GET YOUR FREE QUOTE` renders 948px wide against a 1080px canvas, so the next",
     "         slightly longer call to action runs off the frame. Two lines beat an edge. */",
@@ -33211,6 +33230,10 @@ var scaffoldAdCommand = defineCommand104({
       // Told, not guessed: the voice description is written in the ad's own language
       // and the engine's trait parser reads English only.
       voiceLanguage: voiceLanguageFor(spec.data),
+      // The cast avatar's own voice. An avatar is a face AND a voice; grounding every
+      // frame on their identity sheet while a separately cast voice reads the lines is
+      // how a presenter ends up speaking in somebody else's.
+      ...avatar?.voiceId ? { voiceId: avatar.voiceId } : {},
       ...spec.data.voiceover === false ? { staticTranscriptPath: path24.relative(outDir, transcriptPath) } : {}
     };
     const canvas = scaffoldVideoCanvas(blueprint, adSpecCastElements(spec.data, avatar), opts);
@@ -33264,6 +33287,9 @@ var scaffoldAdCommand = defineCommand104({
         ] : [],
         ...avatarError ? [avatarError] : [],
         ...avatar ? [
+          ...avatar?.voiceId ? [] : [
+            `Avatar \`${handle}\` has no pinned voice, so the ad's lines are read by a voice cast from a description \u2014 it will not sound like them. Give them one (\`baker avatars\` voice) before this ships.`
+          ],
           `Every beat the customer appears in is grounded on \`${handle}\`'s identity sheet, and their subject description was copied into the frames verbatim \u2014 so it is the same face throughout and the same face as the rest of this company's work.`
         ] : [],
         ...presenterIsInEveryShot(spec.data) ? [