@cueframe/kernel 0.2.51 → 0.2.53

package/dist/{chunk-IHWC75AN.js → chunk-5E4YP44R.js}

@@ -2,7 +2,7 @@ import {
   COMPOSE_PLANNER_TOOLS,
   COMPOSE_ROLE_TOOLS,
   dispatchComposeRoleTool
-} from "./chunk-KY3QBG7O.js";
+} from "./chunk-4DM63RTI.js";
 
 // ../authoring/src/agent/buildTools.ts
 import { tool } from "ai";

package/dist/{chunk-MOUWO53S.js → chunk-7AUHC2X7.js}

@@ -1,7 +1,7 @@
 import {
   bodySchemaFor,
   opById
-} from "./chunk-2LUAUF46.js";
+} from "./chunk-46DGDPXR.js";
 
 // ../mcp/src/generated/sceneContract.ts
 function deepFreeze(value) {

package/dist/{chunk-AVTEPAWI.js → chunk-7RGDRDAR.js}

@@ -75,7 +75,7 @@ var BUNDLED_FONT_FACES = Object.freeze({
   },
   "Instrument Serif": {
     slug: "instrument-serif",
-    faces: [{ weight: "400", file: "400" }],
+    faces: [{ weight: "400", file: "400" }, { weight: "400", file: "400-italic", style: "italic" }],
     captionDefaults: { base: 400, emphasis: 400 }
   },
   "National Park": {
@@ -102,19 +102,19 @@ function bundledFaceCoversWeight(face, weight) {
   return weight >= lo && weight <= hi;
 }
 function bundledFontFaceExists(family, weight, style) {
-  if (style !== BUNDLED_FONT_STYLE) return false;
   const bundled = bundledFontFamilyFaces(family);
   if (!bundled) return false;
-  return bundled.faces.some((face) => bundledFaceCoversWeight(face, weight));
+  return bundled.faces.some((face) => (face.style ?? BUNDLED_FONT_STYLE) === style && bundledFaceCoversWeight(face, weight));
 }
 function describeBundledFontWeights(family) {
   const bundled = bundledFontFamilyFaces(family);
   if (!bundled) return "";
-  return bundled.faces.map((face) => face.weight.split(/\s+/).join("-")).join(", ");
+  return [...new Set(bundled.faces.map((face) => face.weight.split(/\s+/).join("-")))].join(", ");
 }
 function unrenderableFaceField(family, weight, style, sealed) {
-  if (bundledFontFamilyFaces(family)) {
-    return style === BUNDLED_FONT_STYLE ? "fontWeight" : "fontStyle";
+  const bundled = bundledFontFamilyFaces(family);
+  if (bundled) {
+    return bundled.faces.some((face) => (face.style ?? BUNDLED_FONT_STYLE) === style) ? "fontWeight" : "fontStyle";
   }
   if (sealed.hasFamily(family)) {
     return sealed.hasWeight(family, weight) ? "fontStyle" : "fontWeight";
@@ -165,7 +165,7 @@ function admissibleFamilyForStack(stack) {
   return resolved.kind === "bundled" ? resolved.family : stack;
 }
 function weightRanges(faces) {
-  return faces.map((face) => {
+  return [...new Map(faces.map((face) => [face.weight, face])).values()].map((face) => {
     const parts = face.weight.split(/\s+/).map(Number);
     const min = parts[0];
     const max = parts.length > 1 ? parts[1] : min;
@@ -188,7 +188,7 @@ function vendoredFontCatalog() {
     return {
       family,
       weights: weightRanges(entry.faces),
-      styles: [BUNDLED_FONT_STYLE],
+      styles: [...new Set(entry.faces.map((face) => face.style ?? BUNDLED_FONT_STYLE))],
       captionDefaults: entry.captionDefaults,
       origin: aliased.has(family) ? "aliased" : "bundled"
     };
@@ -261,12 +261,12 @@ function faceIssue(use) {
     return {
       field: use.field,
       face: { family: use.stack, weight: use.weight, style: use.style },
-      message: `font_not_reproducible \u2014 ${because(use)}: "${use.stack}" names no font file the render bundle carries, and it ships no italic binary at all, so the browser would slant whatever upright face it matched and the render refuses it before capture. Emphasise with emphasisFontWeight, emphasisColor or emphasisScale, or name an uploaded brand font that is sealed in italic.`
+      message: `font_not_reproducible \u2014 ${because(use)}: "${use.stack}" names no font file the render bundle carries, so this stack ships no italic binary to select and the browser would slant whatever upright face it matched and the render refuses it before capture. Emphasise with emphasisFontWeight, emphasisColor or emphasisScale, or name an uploaded brand font that is sealed in italic.`
     };
   }
   if (use.resolved.kind !== "bundled") return null;
   const face = { family: use.resolved.family, weight: use.weight, style: use.style };
-  if (use.style === "italic") {
+  if (use.style === "italic" && !bundledFontFamilyFaces(use.resolved.family).faces.some((face2) => face2.style === "italic")) {
     return {
       field: use.field,
       face,
@@ -274,10 +274,11 @@ function faceIssue(use) {
     };
   }
   if (bundledFontFaceExists(use.resolved.family, use.weight, use.style)) return null;
+  const weightField = use.role === "emphasis" ? "emphasisFontWeight" : "fontWeight";
   return {
-    field: use.field,
+    field: weightField,
     face,
-    message: `font_not_reproducible \u2014 ${because(use)}: ${use.resolved.family} ships ${describeBundledFontWeights(use.resolved.family)}. The render refuses a weight outside that range before capture rather than let the browser substitute a nearby file, so set ${use.field} to a weight the family covers, or pick a family whose files do.`
+    message: `font_not_reproducible \u2014 ${weightField} ${use.weight} (${use.style}) has no file behind it: ${use.resolved.family} ships ${describeBundledFontWeights(use.resolved.family)}. The render refuses a weight outside that range before capture rather than let the browser substitute a nearby file, so set ${weightField} to a weight the family covers, or pick a family whose files do.`
   };
 }
 function captionFaceIssues(style) {
@@ -377,7 +378,10 @@ function captionWeightChoices(stack) {
   );
 }
 function captionStyleChoices(stack) {
-  return captionStackFamily(stack).kind === "unknown" ? ["normal", "italic"] : ["normal"];
+  const resolved = captionStackFamily(stack);
+  if (resolved.kind === "unknown") return ["normal", "italic"];
+  if (resolved.kind !== "bundled") return ["normal"];
+  return [...new Set(bundledFontFamilyFaces(resolved.family).faces.map((face) => face.style ?? BUNDLED_FONT_STYLE))];
 }
 
 // ../composition/src/fonts.ts

package/dist/{chunk-7XZW26U2.js → chunk-AONWKKXK.js}

@@ -31,7 +31,7 @@ import {
 } from "./chunk-CL5Y6V6K.js";
 import {
   BUNDLED_FONT_FACES
-} from "./chunk-AVTEPAWI.js";
+} from "./chunk-7RGDRDAR.js";
 
 // ../execution/component-runtime/src/sceneAdmission.ts
 import { ZodError } from "zod";

package/dist/{chunk-O3SJTX6S.js → chunk-C6ONUJIO.js}

@@ -1,7 +1,3 @@
-import {
-  frostedGlassWipe,
-  spatialPush
-} from "./chunk-QGAWDLYX.js";
 import {
   DashboardPopulate
 } from "./chunk-S5H6GH6M.js";
@@ -11,6 +7,10 @@ import {
 import {
   EASING_BY_INTENT
 } from "./chunk-MFHXCHRI.js";
+import {
+  frostedGlassWipe,
+  spatialPush
+} from "./chunk-QGAWDLYX.js";
 import {
   ACCENT,
   ACCENT_SOFT,
@@ -43,7 +43,7 @@ import {
   makeDefaultScreenTexture,
   resolveCamera,
   useImageTexture
-} from "./chunk-2PU2TSUG.js";
+} from "./chunk-D5E5XFHK.js";
 import {
   DEVICE_GLASS_OPACITY_ENCODED,
   DeviceMesh,

package/dist/{chunk-LO3DH776.js → chunk-CJMUTCJW.js}

@@ -4,10 +4,10 @@ import {
   Format,
   GraphicRole,
   OpaqueId
-} from "./chunk-LZEX2GAY.js";
+} from "./chunk-UFNIB3JL.js";
 import {
   brandKitCaptionFaceIssues
-} from "./chunk-AVTEPAWI.js";
+} from "./chunk-7RGDRDAR.js";
 
 // ../composition/src/validators/composePlan.ts
 import { z } from "zod";

package/dist/{chunk-CZOVEX2D.js → chunk-CR3G2JQP.js}

@@ -24,7 +24,7 @@ import {
 } from "./chunk-W2245F6G.js";
 import {
   FlexLayout
-} from "./chunk-2PU2TSUG.js";
+} from "./chunk-D5E5XFHK.js";
 
 // ../execution/primitives/src/layouts/chat-to-preview-layout/index.tsx
 import { Easing, interpolate, useCurrentFrame, useVideoConfig } from "remotion";

package/dist/{chunk-2PU2TSUG.js → chunk-D5E5XFHK.js}

@@ -34,7 +34,7 @@ import {
 } from "./chunk-CL5Y6V6K.js";
 import {
   resolveCaptionFaces
-} from "./chunk-AVTEPAWI.js";
+} from "./chunk-7RGDRDAR.js";
 
 // ../execution/animate/src/easing/smooth.ts
 var sigmoid = (x) => 1 / (1 + Math.exp(-x));

package/dist/{chunk-6WQHVYJD.js → chunk-G7ZSFSQG.js}

@@ -2,7 +2,7 @@ import {
   activeFaceKeepOutSeries,
   boundingBox,
   resolveBrandRefsDeep
-} from "./chunk-Z6NY6MCL.js";
+} from "./chunk-RMLGMYEJ.js";
 import {
   CueframeError,
   encodeError

package/dist/{chunk-7SXY6NFW.js → chunk-GAACUQAB.js}

@@ -1,9 +1,9 @@
-import {
-  FitText
-} from "./chunk-XHGZ2E77.js";
 import {
   EASING_BY_INTENT
 } from "./chunk-MFHXCHRI.js";
+import {
+  FitText
+} from "./chunk-XHGZ2E77.js";
 import {
   ACCENT,
   ACCENT_STRONG,
@@ -26,7 +26,7 @@ import {
   perspectiveFor,
   resolveCamera,
   useStage
-} from "./chunk-2PU2TSUG.js";
+} from "./chunk-D5E5XFHK.js";
 
 // ../execution/primitives/src/effects/blur/index.tsx
 import React2 from "react";

package/dist/{chunk-4CNJGY7P.js → chunk-LPUQTBKU.js}

@@ -31,7 +31,7 @@ import {
 } from "./chunk-VZ7TYX6Z.js";
 import {
   resolveBrandRef
-} from "./chunk-Z6NY6MCL.js";
+} from "./chunk-RMLGMYEJ.js";
 import {
   SHOT_SCALE_TARGET,
   shotScaleZoom
@@ -64,7 +64,7 @@ import {
   bundledFontFaceExists,
   mergedCaptionFaceIssues,
   unrenderableFaceField
-} from "./chunk-AVTEPAWI.js";
+} from "./chunk-7RGDRDAR.js";
 
 // ../execution/core/src/render/sceneCaptureIdentity.ts
 var INTENT_FIELDS = [
@@ -1226,29 +1226,41 @@ function bindBehindSubjectHosts(overlays, clips) {
     );
   }
 }
-function assertBehindSubjectCaptionCoverage(captionSegments, clips, fps, output, effects = []) {
-  for (const segment of captionSegments ?? []) {
-    for (const word of segment) {
-      const startFrame = Math.round(word.startMs / 1e3 * fps);
-      const endFrame = Math.max(startFrame + 1, Math.round(word.endMs / 1e3 * fps));
-      for (const window of visiblePackedHostWindows(clips, startFrame, endFrame, output, effects)) {
-        if (window.hostClipId) continue;
-        const code = window.candidateClipIds.length > 1 ? "behind_subject_caption_host_ambiguous" : "behind_subject_caption_host_missing";
-        throw renderInputError(
-          code,
-          `Behind-subject caption word '${word.text}' is not fully covered by a unique visible packed host during frames [${window.startFrame}, ${window.endFrame}); an exact matte is required for the visible picture.`,
-          { candidateClipIds: window.candidateClipIds, startFrame: window.startFrame, endFrame: window.endFrame }
-        );
-      }
+function assertBehindSubjectCaptionCoverage(captionSegments, clips, fps, output, effects = [], captionLayouts = []) {
+  const requireCoverage = (label, startMs, endMs) => {
+    const startFrame = Math.round(startMs / 1e3 * fps);
+    const endFrame = Math.max(startFrame + 1, Math.round(endMs / 1e3 * fps));
+    for (const window of visiblePackedHostWindows(clips, startFrame, endFrame, output, effects)) {
+      if (!window.captionVisible || window.hostClipId) continue;
+      const code = window.candidateClipIds.length > 1 ? "behind_subject_caption_host_ambiguous" : "behind_subject_caption_host_missing";
+      throw renderInputError(
+        code,
+        `Behind-subject caption ${label} is not fully covered by a unique visible packed host during frames [${window.startFrame}, ${window.endFrame}); an exact matte is required for the visible picture.`,
+        { candidateClipIds: window.candidateClipIds, startFrame: window.startFrame, endFrame: window.endFrame }
+      );
     }
+  };
+  for (const segment of captionSegments ?? []) {
+    for (const word of segment) requireCoverage(`word '${word.text}'`, word.startMs, word.endMs);
+  }
+  for (const layout of captionLayouts ?? []) {
+    const segment = captionSegments?.[layout.segmentIndex];
+    const first = segment?.[layout.wordStartIndex];
+    const last = segment?.[layout.holdThroughWordIndex ?? layout.wordEndIndex - 1];
+    if (!first || !last) throw new Error("Behind-subject caption layout references missing saved transcript words");
+    requireCoverage(
+      `layout segment ${layout.segmentIndex} words ${layout.wordStartIndex}..${layout.wordEndIndex - 1}`,
+      first.startMs,
+      last.endMs
+    );
   }
 }
 var FULL_RECT = { left: 0, top: 0, right: 1, bottom: 1 };
 function hasClipEffect(clip, effects) {
-  return effects.some((effect) => effect.target.kind === "clip" && effect.trackId === clip.trackId && effect.contentIndex === clip.contentIndex);
+  return clip.hasClipEffects === true || effects.some((effect) => effect.target.kind === "clip" && effect.trackId === clip.trackId && effect.contentIndex === clip.contentIndex);
 }
 function pictureRect(clip, output, effects) {
-  if (clip.cropTrajectory || clip.cropTrajectoryUrl || hasClipEffect(clip, effects)) return void 0;
+  if (clip.cropTrajectory || clip.cropTrajectoryUrl || clip.hasCropIntent || hasClipEffect(clip, effects)) return void 0;
   if (clip.fit !== "contain") return FULL_RECT;
   const sourceWidth = clip.sourceWidth;
   const sourceHeight = clip.sourceHeight;
@@ -1271,7 +1283,7 @@ function coversLowerPictures(upper, lower, output, effects) {
     return upperRect.left <= rect.left + epsilon && upperRect.top <= rect.top + epsilon && upperRect.right + epsilon >= rect.right && upperRect.bottom + epsilon >= rect.bottom;
   });
 }
-function visiblePackedHostWindows(clips, startFrame, endFrame, output, effects = []) {
+function visiblePackedHostWindows(clips, startFrame, endFrame, output, effects = [], hostMode = "packed-only") {
   if (endFrame <= startFrame) return [];
   const boundaries = /* @__PURE__ */ new Set([startFrame, endFrame]);
   for (const clip of clips) {
@@ -1280,6 +1292,13 @@ function visiblePackedHostWindows(clips, startFrame, endFrame, output, effects =
     if (start > startFrame && start < endFrame) boundaries.add(start);
     if (end > startFrame && end < endFrame) boundaries.add(end);
   }
+  for (const effect of effects) {
+    if (effect.target.kind !== "global") continue;
+    const start = effect.startFrame ?? startFrame;
+    const end = start + (effect.durationInFrames ?? endFrame - start);
+    if (start > startFrame && start < endFrame) boundaries.add(start);
+    if (end > startFrame && end < endFrame) boundaries.add(end);
+  }
   const points = [...boundaries].sort((a, b) => a - b);
   const windows = [];
   for (let i = 0; i < points.length - 1; i++) {
@@ -1290,6 +1309,7 @@ function visiblePackedHostWindows(clips, startFrame, endFrame, output, effects =
     );
     const topLayer = Math.max(...active.map((clip) => clip.layerIndex));
     const top = active.filter((clip) => clip.layerIndex === topLayer);
+    const globalEffectActive = effects.some((effect) => effect.target.kind === "global" && (effect.startFrame ?? startFrame) <= start && (effect.startFrame ?? startFrame) + (effect.durationInFrames ?? endFrame - startFrame) >= end);
     const fullFrame = top[0]?.region === void 0 || top[0].region.x === 0 && top[0].region.y === 0 && top[0].region.w === 1 && top[0].region.h === 1;
     const provesOcclusion = top.length === 1 && fullFrame && (active.length === 1 || coversLowerPictures(
       top[0],
@@ -1297,18 +1317,22 @@ function visiblePackedHostWindows(clips, startFrame, endFrame, output, effects =
       output,
       effects
     ));
-    const host = provesOcclusion && top[0].packedUrl ? top[0] : void 0;
-    const unresolvedOverlap = top.length > 1 || top.length === 1 && !!top[0].packedUrl && active.length > 1 && !provesOcclusion;
-    const candidates = unresolvedOverlap ? active.filter((clip) => clip.packedUrl).map((clip) => clip.id) : top.map((clip) => clip.id);
+    const suppressesCaption = provesOcclusion && !globalEffectActive && pictureRect(top[0], output, effects) !== void 0;
+    const captionVisible = !(suppressesCaption && top[0].captionVisibility === "hidden");
+    const candidateHost = (clip) => hostMode === "all-video-candidates" ? clip.mediaType !== "image" : !!clip.packedUrl;
+    const host = provesOcclusion && candidateHost(top[0]) ? top[0] : void 0;
+    const unresolvedOverlap = top.length > 1 || top.length === 1 && active.length > 1 && !provesOcclusion;
+    const candidates = unresolvedOverlap ? active.filter(candidateHost).map((clip) => clip.id) : top.map((clip) => clip.id);
     const previous = windows[windows.length - 1];
-    if (previous && previous.endFrame === start && previous.hostClipId === host?.id && previous.candidateClipIds.join("\0") === candidates.join("\0")) {
+    if (previous && previous.endFrame === start && previous.hostClipId === host?.id && previous.captionVisible === captionVisible && previous.candidateClipIds.join("\0") === candidates.join("\0")) {
       previous.endFrame = end;
     } else {
       windows.push({
         startFrame: start,
         endFrame: end,
         ...host ? { hostClipId: host.id } : {},
-        candidateClipIds: candidates
+        candidateClipIds: candidates,
+        captionVisible
       });
     }
   }
@@ -1455,6 +1479,13 @@ function projectMediaTrack(track, trackIndex, trackCount, fps, srcUrls, trajecto
       );
     }
     if (rendersNoFrames(c)) continue;
+    if (c.captionVisibility !== void 0 && track.kind !== "video") {
+      throw renderInputError(
+        "invalid_composition",
+        `Clip ${c.id}: captionVisibility is supported only on video media clips.`,
+        { trackId: track.id, clipId: c.id, compositionPath: `composition.tracks[${track.id}].contents[${i}].captionVisibility` }
+      );
+    }
     const src = resolveSrcUrl(c.source.mediaId, srcUrls, {
       trackId: track.id,
       clipId: c.id,
@@ -1546,6 +1577,7 @@ function projectMediaTrack(track, trackIndex, trackCount, fps, srcUrls, trajecto
         ...c.animate !== void 0 ? { animate: c.animate } : {},
         ...c.region !== void 0 ? { region: c.region } : {},
         ...c.fit !== void 0 ? { fit: c.fit } : {},
+        ...c.captionVisibility !== void 0 ? { captionVisibility: c.captionVisibility } : {},
         ...s === 0 && c.transitionIn ? {
           transitionIn: mapTransitionType(c.transitionIn.type),
           transitionDurationFrames: Math.max(1, Math.round(c.transitionIn.duration * fps))
@@ -1813,6 +1845,71 @@ function projectEffectTrack(track, trackIndex, trackCount, fps) {
 }
 
 // ../execution/core/src/render/compositionRender.ts
+function captionVisibilityWindowsForComposition(composition, sourceDimensions) {
+  const { fps, width, height } = compositionFormatDimensions(composition.format);
+  const clips = [];
+  const mediaIdByProjectedClipId = {};
+  let endFrame = 0;
+  const trackCount = composition.tracks.length;
+  const effects = composition.tracks.flatMap(
+    (track, index) => track.enabled !== false && track.kind === "effect" ? projectEffectTrack(track, index, trackCount, fps) : []
+  );
+  for (const [trackIndex, track] of composition.tracks.entries()) {
+    if (track.enabled === false || track.kind !== "video" && track.kind !== "image") continue;
+    for (const [contentIndex, item] of track.contents.entries()) {
+      if (!("source" in item)) continue;
+      if (item.source.kind !== "media") {
+        throw renderInputError(
+          "clip_source_track_mismatch",
+          `Clip ${item.id} on video track ${track.id} must use a media source.`,
+          { trackId: track.id, clipId: item.id }
+        );
+      }
+      const spans = projectMediaPlaybackSpans(track.kind === "video" ? {
+        startTime: item.startTime,
+        duration: item.duration,
+        trim: item.source.trim,
+        excludedRanges: item.source.excludedRanges,
+        playbackRate: item.source.playbackRate,
+        fps
+      } : { startTime: item.startTime, duration: item.duration, fps });
+      const dimensions = sourceDimensions[item.source.mediaId];
+      const validDimensions = track.kind === "video" && dimensions && Number.isFinite(dimensions.width) && dimensions.width > 0 && Number.isFinite(dimensions.height) && dimensions.height > 0;
+      for (const [spanIndex, span] of spans.entries()) {
+        if (span.durationInFrames === 0) continue;
+        const id = spans.length === 1 ? item.id : `${item.id}#${spanIndex}`;
+        clips.push({
+          id,
+          trackId: track.id,
+          layerIndex: trackCount - trackIndex,
+          contentIndex,
+          timelineStartFrames: span.timelineStartFrames,
+          durationInFrames: span.durationInFrames,
+          ...track.kind === "image" ? { mediaType: "image" } : {},
+          ...validDimensions ? { sourceWidth: dimensions.width, sourceHeight: dimensions.height } : {},
+          ...item.region ? { region: item.region } : {},
+          ...item.fit ? { fit: item.fit } : {},
+          ...track.kind === "video" && item.captionVisibility ? { captionVisibility: item.captionVisibility } : {},
+          ...item.source.reframe?.segments.length ? { hasCropIntent: true } : {},
+          ...item.effects?.length ? { hasClipEffects: true } : {}
+        });
+        if (track.kind === "video") mediaIdByProjectedClipId[id] = item.source.mediaId;
+        endFrame = Math.max(endFrame, span.timelineStartFrames + span.durationInFrames);
+      }
+    }
+  }
+  for (const segment of composition.captions?.segments ?? []) {
+    for (const word of segment.words) {
+      const startFrame = Math.round(word.startMs / 1e3 * fps);
+      endFrame = Math.max(endFrame, Math.max(startFrame + 1, Math.round(word.endMs / 1e3 * fps)));
+    }
+  }
+  return {
+    fps,
+    windows: visiblePackedHostWindows(clips, 0, endFrame, { width, height }, effects, "all-video-candidates"),
+    mediaIdByProjectedClipId
+  };
+}
 var CAPTION_COLOR_FIELDS = [
   "color",
   "highlightColor",
@@ -1890,6 +1987,11 @@ function assertMergedCaptionFacesDrawable(composition) {
   if (!captions?.segments.some((segment) => segment.words.length > 0)) return;
   const brandCaption = composition.brandKit?.caption;
   const issues = mergedCaptionFaceIssues(captions.style, brandCaption);
+  for (const [index, layout] of (captions.layouts ?? []).entries()) {
+    for (const issue of mergedCaptionFaceIssues({ ...captions.style, ...layout.style }, brandCaption)) {
+      issues.push({ ...issue, path: `captions.layouts[${index}].style.${issue.path.split(".").at(-1)}` });
+    }
+  }
   if (issues.length === 0) return;
   throw renderInputError(
     "font_not_reproducible",
@@ -1903,6 +2005,18 @@ function assertMergedCaptionFacesDrawable(composition) {
 function compositionToRenderInput(composition, srcUrls, trajectoryUrls, coViewUrls, alphaUrls, subCompositions, trajectories, options) {
   const flattened = hasCompoundClips(composition) ? flattenCompoundClips(composition, (id) => subCompositions?.[id]) : composition;
   composition = flattened;
+  for (const track of composition.tracks) {
+    for (const content of track.contents) {
+      if (!("source" in content) || content.captionVisibility === void 0) continue;
+      if (track.kind !== "video" || content.source.kind !== "media") {
+        throw renderInputError(
+          "invalid_composition",
+          `Clip ${content.id}: captionVisibility is supported only on video media clips.`,
+          { trackId: track.id, clipId: content.id, compositionPath: `composition.tracks[${track.id}].contents[${content.id}].captionVisibility` }
+        );
+      }
+    }
+  }
   assertMergedCaptionFacesDrawable(composition);
   const { width, height, fps } = compositionFormatDimensions(composition.format);
   const clips = [];
@@ -1980,9 +2094,9 @@ function compositionToRenderInput(composition, srcUrls, trajectoryUrls, coViewUr
     if (end > maxBackgroundEndFrames) maxBackgroundEndFrames = end;
   }
   const bodyDurationInFrames = maxBackgroundEndFrames > 0 ? Math.min(maxEndFrames, maxBackgroundEndFrames) : maxEndFrames;
-  const captionSegments = composition.captions?.segments.map((seg) => seg.words).filter((words) => words.length > 0);
+  const captionSegments = composition.captions?.segments.map((seg) => seg.words);
   if (composition.captions?.zPlane === "behind-subject") {
-    assertBehindSubjectCaptionCoverage(captionSegments, clips, fps, { width, height }, effects);
+    assertBehindSubjectCaptionCoverage(captionSegments, clips, fps, { width, height }, effects, composition.captions.layouts);
   }
   const brandKit = composition.brandKit;
   const { totalFrames: durationInFrames } = resolveBookendTimeline(
@@ -2004,7 +2118,7 @@ function compositionToRenderInput(composition, srcUrls, trajectoryUrls, coViewUr
     ...mot.exitMs != null ? { exitMs: mot.exitMs } : {}
   } : void 0;
   const baseStyle = composition.captions?.style;
-  const hasCaptions = captionSegments != null && captionSegments.length > 0;
+  const hasCaptions = captionSegments?.some((words) => words.length > 0) ?? false;
   const captionStyle = hasCaptions && (cap || baseStyle || captionMotion) ? {
     ...baseStyle ?? {},
     ...cap?.fontFamily ? { fontFamily: cap.fontFamily } : {},
@@ -2029,6 +2143,10 @@ function compositionToRenderInput(composition, srcUrls, trajectoryUrls, coViewUr
     ...cap?.exitMs != null ? { exitMs: cap.exitMs } : {}
   } : void 0;
   const resolvedCaptionStyle = captionStyle ? resolveCaptionColorRefs(captionStyle, brandKit) : void 0;
+  const captionLayouts = composition.captions?.layouts?.map((layout) => ({
+    ...layout,
+    ...layout.style ? { style: resolveCaptionColorRefs(layout.style, brandKit) } : {}
+  }));
   return {
     clips,
     overlays,
@@ -2040,6 +2158,7 @@ function compositionToRenderInput(composition, srcUrls, trajectoryUrls, coViewUr
     ...audioTracks.length > 0 ? { audioTracks } : {},
     ...effects.length > 0 ? { effects } : {},
     ...hasCaptions ? { captionSegments } : {},
+    ...hasCaptions && captionLayouts?.length ? { captionLayouts } : {},
     ...hasCaptions && composition.captions?.zPlane === "behind-subject" ? { captionsBehind: true } : {},
     ...resolvedCaptionStyle ? { captionStyle: resolvedCaptionStyle } : {},
     ...brandKit ? { brandKit } : {}
@@ -3086,6 +3205,7 @@ export {
   visiblePackedHostWindows,
   playableSpans,
   projectMediaPlaybackSpans,
+  captionVisibilityWindowsForComposition,
   compositionFormatDimensions,
   referencedMediaIds,
   compositionToRenderInput,

package/dist/{chunk-YKXOAGY3.js → chunk-P3DYA5NA.js}

@@ -1,16 +1,16 @@
+import {
+  SCENE_SHOT_RESOURCE_DESCRIPTION,
+  SCENE_SHOT_RESOURCE_URI,
+  sceneShotContract
+} from "./chunk-7AUHC2X7.js";
 import {
   querySchemaShape,
   serializeQuery,
   waitForJob
-} from "./chunk-FSA3UBVR.js";
+} from "./chunk-ZFW35DLG.js";
 import {
   CURATED_TOOLS
 } from "./chunk-4WBSXHQA.js";
-import {
-  SCENE_SHOT_RESOURCE_DESCRIPTION,
-  SCENE_SHOT_RESOURCE_URI,
-  sceneShotContract
-} from "./chunk-MOUWO53S.js";
 import {
   GENERATED_OPERATIONS,
   apiErrorFrom,
@@ -28,10 +28,10 @@ import {
   text,
   textWithData,
   withMedia
-} from "./chunk-2LUAUF46.js";
+} from "./chunk-46DGDPXR.js";
 import {
   CompositionOp
-} from "./chunk-M665ZQYB.js";
+} from "./chunk-VTM7NEJC.js";
 
 // ../mcp/src/prompts.ts
 var skillListCache = null;

package/dist/{chunk-Z6NY6MCL.js → chunk-RMLGMYEJ.js}

@@ -10,7 +10,7 @@ import {
 import {
   captionFontSource,
   effectiveCaptionFamily
-} from "./chunk-AVTEPAWI.js";
+} from "./chunk-7RGDRDAR.js";
 
 // ../composition/src/projection/brandKit.ts
 var LOGO_DEFAULT_PLACEMENT = "br";