@koda-sl/baker-cli 0.281.4-dev.1f1c09c80 → 0.281.5-dev.1f1c09c80

package/dist/cli.js

@@ -32791,6 +32791,24 @@ var AdSpec = z30.object({
    * is all this needs to carry.
    */
   music: z30.union([z30.string().min(1), z30.literal(false)]).optional(),
+  /**
+   * The place the ad keeps coming back to, held still across every shot.
+   *
+   * The person has an identity sheet; nothing held the SETTING. Each shot's start frame is
+   * generated on its own, so three beats in "her living room" came back as three different
+   * living rooms. The engine already builds `location` reference sheets — its own comment
+   * calls them the strongest cross-scene consistency lock it has — and the ad path simply
+   * never emitted one.
+   *
+   * `image` is a real picture of the place, and it is what makes this work at all: a sheet
+   * fuses source material into a turnaround. Fed nothing, it is not a reference sheet, it
+   * is `image_generate` with extra steps — the model registry says exactly that. Generate
+   * the plate once with gpt-image-2 and pass its URL here.
+   */
+  set: z30.object({
+    description: z30.string().min(1).max(400),
+    image: z30.string().min(1).optional()
+  }).optional(),
   /**
    * How the ad closes.
    *
@@ -32938,6 +32956,7 @@ function endCardCta(spec) {
   return declared.toLowerCase().replace(/[.!?]$/, "") === closing.replace(/[.!?]$/, "") ? null : declared;
 }
 var AD_CAST_ID = "customer";
+var AD_SET_ID = "set";
 function castedAvatarOnRoster(cast, roster) {
   if (!cast || cast.avatar || !cast.description) return null;
   const words2 = new Set(
@@ -32953,27 +32972,39 @@ function castedAvatarOnRoster(cast, roster) {
   return null;
 }
 function adSpecCastElements(spec, avatar) {
-  const description = (avatar?.subjectDescription ?? spec.cast?.description)?.trim();
-  if (!description) return [];
   const lastIndex = spec.beats.length - 1;
+  const speakingScenes = spec.beats.map((_, i) => i).filter((i) => !(endCardWanted(spec) && i === lastIndex));
+  const place = spec.set?.description?.trim();
+  const elements = [];
+  if (place && speakingScenes.length > 0) {
+    elements.push({
+      type: "location",
+      label: AD_SET_ID,
+      cast_id: AD_SET_ID,
+      description: place,
+      scenes: speakingScenes,
+      ...spec.set?.image?.trim() ? { source_image: spec.set.image.trim() } : {}
+    });
+  }
+  const description = (avatar?.subjectDescription ?? spec.cast?.description)?.trim();
+  if (!description) return elements;
   const scenes = spec.beats.map((beat, i) => ({ beat, i })).filter(({ beat, i }) => beat.cast !== false && !(endCardWanted(spec) && i === lastIndex)).map(({ i }) => i);
-  if (scenes.length === 0) return [];
+  if (scenes.length === 0) return elements;
   const sheet = avatar?.sheetUrl?.trim();
-  return [
-    {
-      type: "person",
-      label: AD_CAST_ID,
-      // Same id the dialogue names as speaker: that pairing is what lets the engine see
-      // the presenter as present in a scene and route it to native audio.
-      cast_id: AD_CAST_ID,
-      description,
-      scenes,
-      // An avatar's identity sheet IS the turnaround the engine would otherwise render,
-      // so it is handed over as one. Left unmarked, the run paid to generate a sheet OF
-      // a sheet and every frame grounded on the copy instead of the avatar.
-      ...sheet ? { source_image: sheet, source_is_sheet: true } : {}
-    }
-  ];
+  elements.push({
+    type: "person",
+    label: AD_CAST_ID,
+    // Same id the dialogue names as speaker: that pairing is what lets the engine see
+    // the presenter as present in a scene and route it to native audio.
+    cast_id: AD_CAST_ID,
+    description,
+    scenes,
+    // An avatar's identity sheet IS the turnaround the engine would otherwise render,
+    // so it is handed over as one. Left unmarked, the run paid to generate a sheet OF
+    // a sheet and every frame grounded on the copy instead of the avatar.
+    ...sheet ? { source_image: sheet, source_is_sheet: true } : {}
+  });
+  return elements;
 }
 function motionPrompt(beat, toCamera, isClosingCard, place, physics) {
   if (isClosingCard) return BRAND_PLATE;
@@ -33412,6 +33443,7 @@ var scaffoldAdCommand = defineCommand105({
       process.exit(1);
       return;
     }
+    let autoCast = null;
     if (!flagAvatar && !spec.data.cast?.avatar && spec.data.cast?.description) {
       const roster = await readAvatars(
         "/api/avatars",
@@ -33419,18 +33451,22 @@ var scaffoldAdCommand = defineCommand105({
       ).catch(() => null);
       const named = roster ? castedAvatarOnRoster(spec.data.cast, roster.avatars) : null;
       if (named) {
-        writeJson({
-          ok: false,
-          error: {
-            code: "VALIDATION_ERROR",
-            message: `This ad describes its cast in words, but \`${named}\` is already one of this company's avatars \u2014 and the description names them. Described, they are re-invented every shot: the face drifts, their own voice is never used, and they cannot be shown speaking.`,
-            fix: `Cast them instead: put \`"cast": { "avatar": "${named}" }\` in the spec, or pass \`--avatar ${named}\`. Keep a written description only for someone the company does not have.`
-          }
-        });
-        process.exit(1);
-        return;
+        spec.data.cast = { ...spec.data.cast, avatar: named };
+        autoCast = named;
       }
     }
+    if (spec.data.set?.description && !spec.data.set.image?.trim()) {
+      writeJson({
+        ok: false,
+        error: {
+          code: "VALIDATION_ERROR",
+          message: "This ad names a recurring place but gives no picture of it. A location sheet is fused from a real image; with nothing to fuse there is no sheet, and every shot invents the room again.",
+          fix: `Make the plate once, then pass it: \`baker images generate "${spec.data.set.description}" --model openai/gpt-image-2\`, then put its url in the spec as \`"set": { "description": "\u2026", "image": "<url>" }\`. Drop \`set\` entirely if the ad does not keep returning to one place.`
+        }
+      });
+      process.exit(1);
+      return;
+    }
     const handle = flagAvatar || spec.data.cast?.avatar?.trim();
     let avatar = null;
     let avatarError = null;
@@ -33625,6 +33661,9 @@ var scaffoldAdCommand = defineCommand105({
       data: { canvas: outPath, beats: spec.data.beats.length, slug },
       hints: [
         `Wrote ${spec.data.beats.length} beats to ${outPath}. Run it with \`baker canvas run ${outPath}\`.`,
+        ...autoCast ? [
+          `Cast \`${autoCast}\` \u2014 the spec described this person in words, and they are one of this company's avatars. Every shot now grounds on their identity sheet and they speak in their own pinned voice. Write \`"cast": { "avatar": "` + autoCast + '" }` yourself next time; a written description is for someone the company does not have.'
+        ] : [],
         `These beats run about ${blueprint.estimated_duration_s}s. If the ask was longer, add beats \u2014 the length is the sum of the lines, so a 25s ad needs roughly 25s of script and cannot be stretched by holding shots.`,
         ...logoPath && !staged ? [`Could not read the logo at \`${logoPath}\` \u2014 the ad has no mark on screen. Give the path as it appears in src/brand/BRAND.md, relative to the workspace root.`] : [],
         ...staged ? ["The brand mark is drawn by the composition \u2014 top-left throughout, and large on the closing card. It is never generated as a picture, so it cannot come back garbled."] : [],