@vibeframe/cli 0.113.2 → 0.113.4

package/dist/commands/_shared/build-cache.d.ts

@@ -1,4 +1,4 @@
-export type BuildAssetKind = "narration" | "backdrop" | "character" | "video" | "music";
+export type BuildAssetKind = "narration" | "backdrop" | "character" | "keyframe" | "video" | "music";
 export interface CacheAssetDescriptor {
     key: string;
     path: string;
@@ -27,6 +27,16 @@ export declare function characterCacheDescriptor(opts: {
     quality: "standard" | "hd";
     size: string;
 }): CacheAssetDescriptor;
+export declare function keyframeCacheDescriptor(opts: {
+    beatId: string;
+    cue: string;
+    provider: string;
+    quality: "standard" | "hd";
+    size: string;
+    ratio?: string;
+    /** Character sheet paths used as edit references — changing them re-keys the keyframe. */
+    characters?: string[];
+}): CacheAssetDescriptor;
 export declare function imageRatioForSize(size: string | undefined): string;
 export declare function videoCacheDescriptor(opts: {
     beatId: string;
@@ -35,6 +45,11 @@ export declare function videoCacheDescriptor(opts: {
     duration: number | undefined;
     /** Character reference image paths — changing them must invalidate the clip. */
     characters?: string[];
+    /**
+     * Keyframe still path for image-to-video mode — changing the keyframe must
+     * invalidate the clip. Omitted for text/reference-to-video clips.
+     */
+    keyframe?: string;
 }): CacheAssetDescriptor;
 export declare function musicCacheDescriptor(opts: {
     beatId: string;

package/dist/commands/_shared/build-plan.d.ts

@@ -15,6 +15,7 @@ export interface BuildPlanBeat {
     assets: {
         narration: AssetPlan | null;
         backdrop: AssetPlan | null;
+        keyframe: AssetPlan | null;
         video: AssetPlan | null;
         music: AssetPlan | null;
     };
@@ -93,6 +94,7 @@ export interface CreateBuildPlanOptions {
     skipNarration?: boolean;
     skipBackdrop?: boolean;
     skipVideo?: boolean;
+    skipKeyframe?: boolean;
     skipMusic?: boolean;
     ttsProvider?: string;
     voice?: string;

package/dist/commands/_shared/scene-build.d.ts

@@ -81,6 +81,23 @@ export type SceneBuildProgressEvent = {
     type: "backdrop-skipped";
     beatId: string;
     reason: string;
+} | {
+    type: "keyframe-cached";
+    beatId: string;
+    path: string;
+} | {
+    type: "keyframe-generated";
+    beatId: string;
+    path: string;
+    provider: string;
+} | {
+    type: "keyframe-failed";
+    beatId: string;
+    error: string;
+} | {
+    type: "keyframe-skipped";
+    beatId: string;
+    reason: string;
 } | {
     type: "video-cached";
     beatId: string;
@@ -168,6 +185,9 @@ export interface BeatBuildOutcome {
     videoMetadataPath?: string;
     videoFreshness?: AssetFreshness;
     videoSourcePath?: string;
+    /** Keyframe still status when the beat uses keyframe → image-to-video mode. */
+    keyframeStatus?: PrimitiveStatus;
+    keyframePath?: string;
     musicStatus: PrimitiveStatus;
     musicPath?: string;
     musicJobId?: string;
@@ -230,6 +250,8 @@ export interface SceneBuildOptions {
     musicProvider?: BuildMusicProvider;
     /** Skip AI video generation even when beats declare video cues. */
     skipVideo?: boolean;
+    /** Skip keyframe-still generation even when beats declare keyframe cues. */
+    skipKeyframe?: boolean;
     /** Skip music generation even when beats declare music cues. */
     skipMusic?: boolean;
     /** OpenAI image quality — see `vibe generate image --quality`. */

package/dist/commands/_shared/storyboard-edit.d.ts

@@ -1,5 +1,5 @@
 import { type Beat } from "./storyboard-parse.js";
-export declare const STORYBOARD_CUE_KEYS: readonly ["duration", "narration", "backdrop", "video", "motion", "voice", "music", "asset", "characters"];
+export declare const STORYBOARD_CUE_KEYS: readonly ["duration", "narration", "backdrop", "video", "keyframe", "motion", "voice", "music", "asset", "characters"];
 export type StoryboardCueKey = (typeof STORYBOARD_CUE_KEYS)[number];
 export interface StoryboardValidationIssue {
     severity: "error" | "warning";

package/dist/commands/_shared/storyboard-parse.d.ts

@@ -113,6 +113,13 @@ export interface BeatCues {
      * reference-to-video generation. A single name or a list.
      */
     characters?: string | string[];
+    /**
+     * Keyframe still prompt. When set, the build generates a keyframe image for
+     * this beat (using the character sheet as an edit reference when `characters`
+     * is also present) and runs Seedance image-to-video on it. The `video` cue,
+     * if present, supplies the motion prompt; otherwise this text is used.
+     */
+    keyframe?: string;
     [key: string]: unknown;
 }
 /** Normalize a beat's `characters` cue to a clean string[] (single or list). */

package/dist/index.js

@@ -31869,6 +31869,7 @@ var init_storyboard_edit = __esm({
       "narration",
       "backdrop",
       "video",
+      "keyframe",
       "motion",
       "voice",
       "music",
@@ -31878,7 +31879,7 @@ var init_storyboard_edit = __esm({
     HEADING_RE2 = /^##\s+(.+?)\s*$/gm;
     LEADING_CUE_RE = /^(\s*)```ya?ml\s*\n([\s\S]*?)\n```\s*(?:\n|$)/;
     ALLOWED_CUE_KEYS = new Set(STORYBOARD_CUE_KEYS);
-    STRING_CUE_KEYS = /* @__PURE__ */ new Set(["narration", "backdrop", "video", "motion", "voice", "music", "asset"]);
+    STRING_CUE_KEYS = /* @__PURE__ */ new Set(["narration", "backdrop", "video", "keyframe", "motion", "voice", "music", "asset"]);
     MAX_RECOMMENDED_BEAT_SEC = 15;
   }
 });
@@ -60422,6 +60423,7 @@ var init_build_asset_reference = __esm({
       narration: [".mp3", ".wav", ".m4a"],
       backdrop: [".png", ".jpg", ".jpeg", ".webp"],
       character: [".png", ".jpg", ".jpeg", ".webp"],
+      keyframe: [".png", ".jpg", ".jpeg", ".webp"],
       video: [".mp4", ".mov", ".webm"],
       music: [".mp3", ".wav", ".m4a"]
     };
@@ -60469,6 +60471,18 @@ function characterCacheDescriptor(opts) {
     ext: "png"
   });
 }
+function keyframeCacheDescriptor(opts) {
+  return cacheAssetDescriptor("keyframe", {
+    beatId: opts.beatId,
+    cue: opts.cue,
+    provider: opts.provider,
+    quality: opts.quality,
+    size: opts.size,
+    ratio: opts.ratio,
+    characters: opts.characters && opts.characters.length > 0 ? opts.characters : void 0,
+    ext: "png"
+  });
+}
 function imageRatioForSize(size) {
   switch (size) {
     case "1024x1024":
@@ -60490,6 +60504,8 @@ function videoCacheDescriptor(opts) {
     ratio: "16:9",
     // Omit when empty so existing (character-less) clips keep their cache key.
     characters: opts.characters && opts.characters.length > 0 ? opts.characters : void 0,
+    // Omit when absent so existing (non-keyframe) clips keep their cache key.
+    keyframe: opts.keyframe || void 0,
     ext: "mp4"
   });
 }
@@ -60640,6 +60656,7 @@ async function createBuildPlan(opts) {
     const backdropPrompt = stringOrUndefined3(cue.backdrop);
     const augmentedBackdropPrompt = backdropPrompt ? augmentBackdropPrompt(backdropPrompt) : null;
     const videoPrompt = stringOrUndefined3(cue.video);
+    const keyframePrompt = stringOrUndefined3(cue.keyframe);
     const musicPrompt = stringOrUndefined3(cue.music);
     const genericReference = resolveGenericAssetReference(projectDir, cue.asset);
     const narrationReference = resolveTypedAssetReference(projectDir, "narration", cue.narration);
@@ -60648,7 +60665,7 @@ async function createBuildPlan(opts) {
     const musicReference = resolveTypedAssetReference(projectDir, "music", cue.music) ?? (!musicPrompt && genericReference?.kind === "music" ? genericReference : null);
     const narrationCue = narrationText ?? narrationReference?.raw;
     const backdropCue = backdropPrompt ?? backdropReference?.raw;
-    const videoCue = videoPrompt ?? videoReference?.raw;
+    const videoCue = videoPrompt ?? videoReference?.raw ?? keyframePrompt;
     const musicCue = musicPrompt ?? musicReference?.raw;
     const narrationCost = resolved.narration.resolved === "elevenlabs" ? ELEVENLABS_NARRATION_COST_USD : resolved.narration.resolved === "openai" ? OPENAI_NARRATION_COST_USD : 0;
     const narrationCache = narrationText && !narrationReference ? narrationCacheDescriptor({
@@ -60666,11 +60683,20 @@ async function createBuildPlan(opts) {
       size: imageSize2,
       ratio: imageRatio
     }) : null;
-    const videoCache = videoPrompt && !videoReference ? videoCacheDescriptor({
+    const videoCache = (videoPrompt || keyframePrompt) && !videoReference ? videoCacheDescriptor({
       beatId: beat.id,
-      cue: videoPrompt,
+      cue: videoPrompt ?? keyframePrompt,
       provider: resolved.video.resolved,
-      duration: beat.duration
+      duration: beat.duration,
+      keyframe: keyframePrompt
+    }) : null;
+    const keyframeCache = keyframePrompt ? keyframeCacheDescriptor({
+      beatId: beat.id,
+      cue: keyframePrompt,
+      provider: resolved.image.resolved,
+      quality: imageQuality,
+      size: imageSize2,
+      ratio: imageRatio
     }) : null;
     const musicCache = musicPrompt && !musicReference ? musicCacheDescriptor({
       beatId: beat.id,
@@ -60707,6 +60733,19 @@ async function createBuildPlan(opts) {
       cost: backdropCostUsd(imageQuality),
       active: includeAssets
     }) : null;
+    const keyframe = keyframePrompt && !opts.skipKeyframe ? assetPlan({
+      kind: "keyframe",
+      beatId: beat.id,
+      cue: keyframePrompt,
+      provider: resolved.image.resolved,
+      path: `assets/keyframe-${beat.id}.png`,
+      cache: keyframeCache,
+      reference: null,
+      projectDir,
+      force: opts.force,
+      cost: backdropCostUsd(imageQuality),
+      active: includeAssets
+    }) : null;
     const video = videoCue && !opts.skipVideo ? assetPlan({
       kind: "video",
       beatId: beat.id,
@@ -60739,7 +60778,7 @@ async function createBuildPlan(opts) {
     }) : null;
     const compositionPath = `compositions/scene-${beat.id}.html`;
     const compositionExists = existsSync47(join23(projectDir, compositionPath));
-    for (const asset of [narration, backdrop, video, music]) {
+    for (const asset of [narration, backdrop, keyframe, video, music]) {
       if (!asset) continue;
       if (asset.referenceError) {
         missing.add("assets");
@@ -60764,6 +60803,8 @@ async function createBuildPlan(opts) {
     if (narration?.willGenerate) noteProviderNeed(resolved.narration, "Narration");
     if (backdrop?.willGenerate && isSupportedBuildImageProvider(resolved.image.resolved))
       noteProviderNeed(resolved.image, "Backdrop generation");
+    if (keyframe?.willGenerate && isSupportedBuildImageProvider(resolved.image.resolved))
+      noteProviderNeed(resolved.image, "Keyframe generation");
     if (video?.willGenerate) noteProviderNeed(resolved.video, "Video generation");
     if (music?.willGenerate) noteProviderNeed(resolved.music, "Music generation");
     if (!compositionExists) missing.add("compositions");
@@ -60777,7 +60818,7 @@ async function createBuildPlan(opts) {
       heading: beat.heading,
       durationSec: beat.duration ?? null,
       cues: cue,
-      assets: { narration, backdrop, video, music },
+      assets: { narration, backdrop, keyframe, video, music },
       composition: {
         path: compositionPath,
         exists: compositionExists
@@ -60993,6 +61034,7 @@ function providerResolutionsForPlan(resolved, beats, opts, includeAssets) {
   return [
     includeAssets && !opts.skipNarration && needsProvider("narration") ? resolved.narration : null,
     includeAssets && !opts.skipBackdrop && needsProvider("backdrop") ? resolved.image : null,
+    includeAssets && !opts.skipKeyframe && needsProvider("keyframe") ? resolved.image : null,
     includeAssets && !opts.skipVideo && needsProvider("video") ? resolved.video : null,
     includeAssets && !opts.skipMusic && needsProvider("music") ? resolved.music : null,
     resolved.composer ?? null
@@ -62062,6 +62104,7 @@ async function executeSceneBuild(opts) {
     skipNarration: opts.skipNarration,
     skipBackdrop: opts.skipBackdrop,
     skipVideo: opts.skipVideo,
+    skipKeyframe: opts.skipKeyframe,
     skipMusic: opts.skipMusic,
     ttsProvider: opts.ttsProvider,
     voice: opts.voice,
@@ -62247,6 +62290,7 @@ async function executeSceneBuild(opts) {
         skipNarration: opts.skipNarration ?? false,
         skipBackdrop: opts.skipBackdrop ?? false,
         skipVideo: opts.skipVideo ?? false,
+        skipKeyframe: opts.skipKeyframe ?? false,
         skipMusic: opts.skipMusic ?? false,
         force: opts.force ?? false,
         onProgress,
@@ -62623,10 +62667,11 @@ async function executeSceneBuild(opts) {
   });
 }
 async function buildBeatPrimitives(beat, ctx) {
+  const keyframe = ctx.skipKeyframe ? await skipped("keyframe", beat.id, "--skip-keyframe", ctx) : await dispatchKeyframe(beat, ctx);
   const [narration, backdrop, video, music] = await Promise.all([
     ctx.skipNarration ? skipped("narration", beat.id, "--skip-narration", ctx) : dispatchNarration(beat, ctx),
     ctx.skipBackdrop ? skipped("backdrop", beat.id, "--skip-backdrop", ctx) : dispatchBackdrop(beat, ctx),
-    ctx.skipVideo ? skipped("video", beat.id, "--skip-video", ctx) : dispatchVideo(beat, ctx),
+    ctx.skipVideo ? skipped("video", beat.id, "--skip-video", ctx) : dispatchVideo(beat, ctx, keyframe),
     ctx.skipMusic ? skipped("music", beat.id, "--skip-music", ctx) : dispatchMusic(beat, ctx)
   ]);
   return {
@@ -62670,6 +62715,8 @@ async function buildBeatPrimitives(beat, ctx) {
       videoMetadataPath: video.metadataPath,
       videoFreshness: video.freshness,
       videoSourcePath: video.sourcePath,
+      keyframeStatus: keyframe.status,
+      keyframePath: keyframe.path,
       musicStatus: music.status,
       musicPath: music.path,
       musicJobId: music.job?.id,
@@ -63114,6 +63161,112 @@ async function generateBackdropImage(prompt3, ctx, ratio) {
   });
   return imageBufferFromResult(result);
 }
+async function editKeyframeImage(prompt3, sheetBuffers, ctx, ratio) {
+  loadSceneBuildEnv(ctx.projectDir);
+  const keyInfo = imageProviderKeyInfo(ctx.imageProvider);
+  const apiKey = await getApiKeyFromConfig(keyInfo.configKey, { cwd: ctx.projectDir }) ?? process.env[keyInfo.envVar] ?? "";
+  if (!apiKey) {
+    return {
+      success: false,
+      error: `${keyInfo.envVar} not set \u2014 cannot generate keyframe with ${ctx.imageProvider}`
+    };
+  }
+  if (ctx.imageProvider === "openai") {
+    const provider2 = new OpenAIImageProvider();
+    await provider2.initialize({ apiKey });
+    const result2 = await provider2.editImage(sheetBuffers, prompt3);
+    return imageBufferFromResult(result2);
+  }
+  if (ctx.imageProvider === "gemini") {
+    const provider2 = new GeminiProvider();
+    await provider2.initialize({ apiKey });
+    const result2 = await provider2.editImage(sheetBuffers, prompt3, {
+      model: "flash",
+      aspectRatio: ratio
+    });
+    return imageBufferFromResult(result2);
+  }
+  const provider = new GrokProvider();
+  await provider.initialize({ apiKey });
+  const result = await provider.editImage(sheetBuffers[0], prompt3, { aspectRatio: ratio });
+  return imageBufferFromResult(result);
+}
+async function ensureKeyframe(beat, ctx, keyframeCue, characterRefsRel) {
+  const rel = `assets/keyframe-${beat.id}.png`;
+  const abs = join25(ctx.projectDir, rel);
+  const size = ctx.imageSize ?? "1536x1024";
+  const ratio = imageRatioForSize(size);
+  const metadataOptions = { quality: ctx.imageQuality, size, ratio };
+  const cache = keyframeCacheDescriptor({
+    beatId: beat.id,
+    cue: keyframeCue,
+    provider: ctx.imageProvider,
+    quality: ctx.imageQuality,
+    size,
+    ratio,
+    characters: characterRefsRel
+  });
+  if (existsSync49(abs) && !ctx.force) {
+    const metadata = readAssetMetadata(ctx.projectDir, "keyframe", beat.id);
+    if (!metadata || isFreshCanonicalAsset({
+      projectDir: ctx.projectDir,
+      kind: "keyframe",
+      beatId: beat.id,
+      cue: keyframeCue,
+      provider: ctx.imageProvider,
+      options: metadataOptions,
+      cacheKey: cache.key
+    })) {
+      return { status: "cached", rel };
+    }
+  }
+  const cacheAbs = join25(ctx.projectDir, cache.path);
+  if (existsSync49(cacheAbs) && !ctx.force) {
+    await mkdir26(dirname24(abs), { recursive: true });
+    await copyFile5(cacheAbs, abs);
+    await writeAssetMetadata({
+      projectDir: ctx.projectDir,
+      kind: "keyframe",
+      beatId: beat.id,
+      cue: keyframeCue,
+      provider: ctx.imageProvider,
+      options: metadataOptions,
+      cacheKey: cache.key,
+      canonicalPath: rel,
+      cachePath: cache.path
+    });
+    return { status: "cached", rel };
+  }
+  const sheetBuffers = [];
+  for (const relPath of characterRefsRel) {
+    const sheetAbs = join25(ctx.projectDir, relPath);
+    if (existsSync49(sheetAbs)) sheetBuffers.push(await readFile46(sheetAbs));
+  }
+  const imgCtx = {
+    projectDir: ctx.projectDir,
+    imageProvider: ctx.imageProvider,
+    imageSize: size,
+    imageQuality: ctx.imageQuality
+  };
+  const generated = sheetBuffers.length > 0 ? await editKeyframeImage(keyframeCue, sheetBuffers, imgCtx, ratio) : await generateBackdropImage(keyframeCue, imgCtx, ratio);
+  if (!generated.success) return { status: "failed", error: generated.error };
+  await mkdir26(dirname24(abs), { recursive: true });
+  await writeFile50(abs, generated.buffer);
+  await mkdir26(dirname24(cacheAbs), { recursive: true });
+  await writeFile50(cacheAbs, generated.buffer);
+  await writeAssetMetadata({
+    projectDir: ctx.projectDir,
+    kind: "keyframe",
+    beatId: beat.id,
+    cue: keyframeCue,
+    provider: ctx.imageProvider,
+    options: metadataOptions,
+    cacheKey: cache.key,
+    canonicalPath: rel,
+    cachePath: cache.path
+  });
+  return { status: "generated", rel };
+}
 async function imageBufferFromResult(result) {
   const image = result.images?.[0];
   if (!result.success || !image) {
@@ -63134,10 +63287,25 @@ function imageProviderKeyInfo(provider) {
   if (provider === "grok") return { configKey: "xai", envVar: "XAI_API_KEY" };
   return { configKey: "openai", envVar: "OPENAI_API_KEY" };
 }
-async function dispatchVideo(beat, ctx) {
+async function dispatchKeyframe(beat, ctx) {
+  const keyframeCue = stringOrUndefined4(beat.cues?.keyframe);
+  if (!keyframeCue) return { status: "no-cue" };
+  const characterRefsRel = beatCharacterNames(beat.cues).map((name) => ctx.characterPaths?.get(name)).filter((p) => typeof p === "string");
+  const kf = await ensureKeyframe(beat, ctx, keyframeCue, characterRefsRel);
+  if (kf.status === "failed") {
+    ctx.onProgress({ type: "keyframe-failed", beatId: beat.id, error: kf.error });
+    return { status: "failed", error: kf.error };
+  }
+  ctx.onProgress(
+    kf.status === "cached" ? { type: "keyframe-cached", beatId: beat.id, path: kf.rel } : { type: "keyframe-generated", beatId: beat.id, path: kf.rel, provider: ctx.imageProvider }
+  );
+  return { status: kf.status, path: kf.rel, provider: ctx.imageProvider };
+}
+async function dispatchVideo(beat, ctx, keyframe) {
   const reference = assetReferenceForBeat(ctx.projectDir, "video", beat);
   if (reference) return referencePrimitiveOutcome("video", beat, ctx, reference);
-  const prompt3 = stringOrUndefined4(beat.cues?.video);
+  const keyframeCue = stringOrUndefined4(beat.cues?.keyframe);
+  const prompt3 = stringOrUndefined4(beat.cues?.video) ?? keyframeCue;
   if (!prompt3) return { status: "no-cue" };
   const characterRefsRel = beatCharacterNames(beat.cues).map((name) => ctx.characterPaths?.get(name)).filter((p) => typeof p === "string");
   const characterRefsAbs = characterRefsRel.map((p) => join25(ctx.projectDir, p));
@@ -63148,7 +63316,10 @@ async function dispatchVideo(beat, ctx) {
     cue: prompt3,
     provider: ctx.videoProvider,
     duration: normalizeVideoDuration(beat.duration),
-    characters: characterRefsRel
+    // Keyframe mode keys the clip to the keyframe prompt; refs feed the keyframe,
+    // not the clip directly.
+    characters: keyframeCue ? void 0 : characterRefsRel,
+    keyframe: keyframeCue
   });
   const metadataPath = assetMetadataPath("video", beat.id);
   if (existsSync49(abs) && !ctx.force) {
@@ -63200,6 +63371,13 @@ async function dispatchVideo(beat, ctx) {
       freshness: "fresh"
     };
   }
+  const keyframeImageRel = keyframeCue && (keyframe.status === "generated" || keyframe.status === "cached") ? keyframe.path : void 0;
+  if (keyframeCue && !keyframeImageRel) {
+    const reason = keyframe.status === "failed" ? `keyframe unavailable: ${keyframe.error ?? "generation failed"}` : "keyframe not generated (build without --skip-keyframe first)";
+    ctx.onProgress({ type: "video-skipped", beatId: beat.id, reason });
+    return { status: "skipped" };
+  }
+  const keyframeImageAbs = keyframeImageRel ? join25(ctx.projectDir, keyframeImageRel) : void 0;
   loadSceneBuildEnv(ctx.projectDir);
   const result = await executeVideoGenerate({
     prompt: prompt3,
@@ -63208,7 +63386,10 @@ async function dispatchVideo(beat, ctx) {
     ratio: "16:9",
     output: abs,
     wait: false,
-    refImages: characterRefsAbs.length > 0 ? characterRefsAbs : void 0,
+    // Keyframe mode → single init frame (image-to-video); otherwise character
+    // reference-to-video.
+    image: keyframeImageAbs,
+    refImages: keyframeImageAbs ? void 0 : characterRefsAbs.length > 0 ? characterRefsAbs : void 0,
     apiKey: await apiKeyForVideoProvider(ctx.videoProvider, ctx.projectDir)
   });
   if (!result.success || !result.taskId) {
@@ -63250,7 +63431,7 @@ async function dispatchVideo(beat, ctx) {
     jobType: "generate-video",
     provider: result.provider ?? ctx.videoProvider,
     providerTaskId: result.taskId,
-    providerTaskType: "text2video",
+    providerTaskType: keyframeImageAbs ? "image2video" : "text2video",
     status: "running",
     projectDir: ctx.projectDir,
     workingDirectory: ctx.projectDir,
@@ -63577,6 +63758,7 @@ function estimateActualAssetCost(outcomes, imageQuality) {
   for (const outcome of outcomes) {
     if (outcome.narrationStatus === "generated") cost += 0.05;
     if (outcome.backdropStatus === "generated") cost += backdropCostUsd(imageQuality);
+    if (outcome.keyframeStatus === "generated") cost += backdropCostUsd(imageQuality);
     if (outcome.videoStatus === "generated" || outcome.videoStatus === "pending") {
       cost += outcome.videoProvider === "seedance" || outcome.videoProvider === "fal" ? estimateSeedanceVideoCostUsd({
         durationSec: normalizeVideoDuration(outcome.sceneDurationSec),
@@ -74662,7 +74844,7 @@ init_source();
 // package.json
 var package_default = {
   name: "@vibeframe/cli",
-  version: "0.113.2",
+  version: "0.113.4",
   description: "VibeFrame CLI - agentic video workflow runtime",
   private: false,
   license: "MIT",
@@ -84948,7 +85130,10 @@ var VALID_STAGES2 = ["assets", "compose", "sync", "render", "all"];
 var VALID_IMAGE_PROVIDERS2 = ["openai", "gemini", "grok"];
 var VALID_VIDEO_PROVIDERS2 = ["seedance", "grok", "kling", "runway", "veo"];
 var VALID_MUSIC_PROVIDERS2 = ["elevenlabs", "replicate"];
-var buildCommand = new Command("build").description("Build a VibeFrame video project from STORYBOARD.md").argument("[project-dir]", "Video project directory", ".").option("--stage <stage>", `Build stage: ${VALID_STAGES2.join("|")}`, "all").option("--beat <id>", "Restrict asset/compose work to one beat id").option("--mode <mode>", "Build mode: agent|batch|auto", "auto").option("--effort <level>", "Compose effort tier (batch mode only): low|medium|high", "medium").option("--composer <provider>", "Batch composer: claude|openai|gemini").option("--max-cost <usd>", "Fail before provider spend when estimated cost exceeds this USD cap").option("--skip-narration", "Don't dispatch TTS even when beats declare narration cues").option("--skip-backdrop", "Don't dispatch image-gen even when beats declare backdrop cues").option("--skip-video", "Don't dispatch video generation even when beats declare video cues").option("--skip-music", "Don't dispatch music generation even when beats declare music cues").option("--skip-render", "Compose only \u2014 don't render to MP4").option("--tts <provider>", "TTS provider: auto|elevenlabs|openai|kokoro").option("--voice <id>", "Voice id").option("--image-provider <name>", `Image provider: ${VALID_IMAGE_PROVIDERS2.join("|")}`).option("--video-provider <name>", `Video provider: ${VALID_VIDEO_PROVIDERS2.join("|")}`).option("--music-provider <name>", `Music provider: ${VALID_MUSIC_PROVIDERS2.join("|")}`).option("--quality <q>", "Image quality: standard|hd", "hd").option("--image-size <s>", "Image size: 1024x1024|1536x1024|1024x1536", "1536x1024").option("--force", "Re-dispatch primitives even when assets already exist").option("--dry-run", "Preview parameters without dispatching").addHelpText(
+var buildCommand = new Command("build").description("Build a VibeFrame video project from STORYBOARD.md").argument("[project-dir]", "Video project directory", ".").option("--stage <stage>", `Build stage: ${VALID_STAGES2.join("|")}`, "all").option("--beat <id>", "Restrict asset/compose work to one beat id").option("--mode <mode>", "Build mode: agent|batch|auto", "auto").option("--effort <level>", "Compose effort tier (batch mode only): low|medium|high", "medium").option("--composer <provider>", "Batch composer: claude|openai|gemini").option("--max-cost <usd>", "Fail before provider spend when estimated cost exceeds this USD cap").option("--skip-narration", "Don't dispatch TTS even when beats declare narration cues").option("--skip-backdrop", "Don't dispatch image-gen even when beats declare backdrop cues").option("--skip-video", "Don't dispatch video generation even when beats declare video cues").option(
+  "--skip-keyframe",
+  "Don't generate keyframe stills (review keyframes first with --skip-video, then build)"
+).option("--skip-music", "Don't dispatch music generation even when beats declare music cues").option("--skip-render", "Compose only \u2014 don't render to MP4").option("--tts <provider>", "TTS provider: auto|elevenlabs|openai|kokoro").option("--voice <id>", "Voice id").option("--image-provider <name>", `Image provider: ${VALID_IMAGE_PROVIDERS2.join("|")}`).option("--video-provider <name>", `Video provider: ${VALID_VIDEO_PROVIDERS2.join("|")}`).option("--music-provider <name>", `Music provider: ${VALID_MUSIC_PROVIDERS2.join("|")}`).option("--quality <q>", "Image quality: standard|hd", "hd").option("--image-size <s>", "Image size: 1024x1024|1536x1024|1024x1536", "1536x1024").option("--force", "Re-dispatch primitives even when assets already exist").option("--dry-run", "Preview parameters without dispatching").addHelpText(
   "after",
   `
 Examples:
@@ -85007,6 +85192,7 @@ Advanced equivalent: \`vibe scene build\`.`
     skipNarration: options.skipNarration ?? false,
     skipBackdrop: options.skipBackdrop ?? false,
     skipVideo: options.skipVideo ?? false,
+    skipKeyframe: options.skipKeyframe ?? false,
     skipMusic: options.skipMusic ?? false,
     skipRender: options.skipRender ?? false,
     ttsProvider: options.tts,
@@ -85027,6 +85213,7 @@ Advanced equivalent: \`vibe scene build\`.`
       skipNarration: options.skipNarration,
       skipBackdrop: options.skipBackdrop,
       skipVideo: options.skipVideo,
+      skipKeyframe: options.skipKeyframe,
       skipMusic: options.skipMusic,
       ttsProvider: options.tts,
       voice: options.voice,
@@ -85117,6 +85304,7 @@ Advanced equivalent: \`vibe scene build\`.`
     skipNarration: options.skipNarration,
     skipBackdrop: options.skipBackdrop,
     skipVideo: options.skipVideo,
+    skipKeyframe: options.skipKeyframe,
     skipMusic: options.skipMusic,
     skipRender: options.skipRender,
     ttsProvider: options.tts,

package/dist/tools/manifest/index.js

@@ -32480,6 +32480,7 @@ var init_storyboard_edit = __esm({
       "narration",
       "backdrop",
       "video",
+      "keyframe",
       "motion",
       "voice",
       "music",
@@ -32489,7 +32490,7 @@ var init_storyboard_edit = __esm({
     HEADING_RE2 = /^##\s+(.+?)\s*$/gm;
     LEADING_CUE_RE = /^(\s*)```ya?ml\s*\n([\s\S]*?)\n```\s*(?:\n|$)/;
     ALLOWED_CUE_KEYS = new Set(STORYBOARD_CUE_KEYS);
-    STRING_CUE_KEYS = /* @__PURE__ */ new Set(["narration", "backdrop", "video", "motion", "voice", "music", "asset"]);
+    STRING_CUE_KEYS = /* @__PURE__ */ new Set(["narration", "backdrop", "video", "keyframe", "motion", "voice", "music", "asset"]);
     MAX_RECOMMENDED_BEAT_SEC = 15;
   }
 });
@@ -33411,6 +33412,7 @@ var init_build_asset_reference = __esm({
       narration: [".mp3", ".wav", ".m4a"],
       backdrop: [".png", ".jpg", ".jpeg", ".webp"],
       character: [".png", ".jpg", ".jpeg", ".webp"],
+      keyframe: [".png", ".jpg", ".jpeg", ".webp"],
       video: [".mp4", ".mov", ".webm"],
       music: [".mp3", ".wav", ".m4a"]
     };
@@ -33458,6 +33460,18 @@ function characterCacheDescriptor(opts) {
     ext: "png"
   });
 }
+function keyframeCacheDescriptor(opts) {
+  return cacheAssetDescriptor("keyframe", {
+    beatId: opts.beatId,
+    cue: opts.cue,
+    provider: opts.provider,
+    quality: opts.quality,
+    size: opts.size,
+    ratio: opts.ratio,
+    characters: opts.characters && opts.characters.length > 0 ? opts.characters : void 0,
+    ext: "png"
+  });
+}
 function imageRatioForSize(size) {
   switch (size) {
     case "1024x1024":
@@ -33479,6 +33493,8 @@ function videoCacheDescriptor(opts) {
     ratio: "16:9",
     // Omit when empty so existing (character-less) clips keep their cache key.
     characters: opts.characters && opts.characters.length > 0 ? opts.characters : void 0,
+    // Omit when absent so existing (non-keyframe) clips keep their cache key.
+    keyframe: opts.keyframe || void 0,
     ext: "mp4"
   });
 }
@@ -33629,6 +33645,7 @@ async function createBuildPlan(opts) {
     const backdropPrompt = stringOrUndefined3(cue.backdrop);
     const augmentedBackdropPrompt = backdropPrompt ? augmentBackdropPrompt(backdropPrompt) : null;
     const videoPrompt = stringOrUndefined3(cue.video);
+    const keyframePrompt = stringOrUndefined3(cue.keyframe);
     const musicPrompt = stringOrUndefined3(cue.music);
     const genericReference = resolveGenericAssetReference(projectDir, cue.asset);
     const narrationReference = resolveTypedAssetReference(projectDir, "narration", cue.narration);
@@ -33637,7 +33654,7 @@ async function createBuildPlan(opts) {
     const musicReference = resolveTypedAssetReference(projectDir, "music", cue.music) ?? (!musicPrompt && genericReference?.kind === "music" ? genericReference : null);
     const narrationCue = narrationText ?? narrationReference?.raw;
     const backdropCue = backdropPrompt ?? backdropReference?.raw;
-    const videoCue = videoPrompt ?? videoReference?.raw;
+    const videoCue = videoPrompt ?? videoReference?.raw ?? keyframePrompt;
     const musicCue = musicPrompt ?? musicReference?.raw;
     const narrationCost = resolved.narration.resolved === "elevenlabs" ? ELEVENLABS_NARRATION_COST_USD : resolved.narration.resolved === "openai" ? OPENAI_NARRATION_COST_USD : 0;
     const narrationCache = narrationText && !narrationReference ? narrationCacheDescriptor({
@@ -33655,11 +33672,20 @@ async function createBuildPlan(opts) {
       size: imageSize2,
       ratio: imageRatio
     }) : null;
-    const videoCache = videoPrompt && !videoReference ? videoCacheDescriptor({
+    const videoCache = (videoPrompt || keyframePrompt) && !videoReference ? videoCacheDescriptor({
       beatId: beat.id,
-      cue: videoPrompt,
+      cue: videoPrompt ?? keyframePrompt,
       provider: resolved.video.resolved,
-      duration: beat.duration
+      duration: beat.duration,
+      keyframe: keyframePrompt
+    }) : null;
+    const keyframeCache = keyframePrompt ? keyframeCacheDescriptor({
+      beatId: beat.id,
+      cue: keyframePrompt,
+      provider: resolved.image.resolved,
+      quality: imageQuality,
+      size: imageSize2,
+      ratio: imageRatio
     }) : null;
     const musicCache = musicPrompt && !musicReference ? musicCacheDescriptor({
       beatId: beat.id,
@@ -33696,6 +33722,19 @@ async function createBuildPlan(opts) {
       cost: backdropCostUsd(imageQuality),
       active: includeAssets
     }) : null;
+    const keyframe = keyframePrompt && !opts.skipKeyframe ? assetPlan({
+      kind: "keyframe",
+      beatId: beat.id,
+      cue: keyframePrompt,
+      provider: resolved.image.resolved,
+      path: `assets/keyframe-${beat.id}.png`,
+      cache: keyframeCache,
+      reference: null,
+      projectDir,
+      force: opts.force,
+      cost: backdropCostUsd(imageQuality),
+      active: includeAssets
+    }) : null;
     const video = videoCue && !opts.skipVideo ? assetPlan({
       kind: "video",
       beatId: beat.id,
@@ -33728,7 +33767,7 @@ async function createBuildPlan(opts) {
     }) : null;
     const compositionPath = `compositions/scene-${beat.id}.html`;
     const compositionExists = existsSync17(join15(projectDir, compositionPath));
-    for (const asset of [narration, backdrop, video, music]) {
+    for (const asset of [narration, backdrop, keyframe, video, music]) {
       if (!asset) continue;
       if (asset.referenceError) {
         missing.add("assets");
@@ -33753,6 +33792,8 @@ async function createBuildPlan(opts) {
     if (narration?.willGenerate) noteProviderNeed(resolved.narration, "Narration");
     if (backdrop?.willGenerate && isSupportedBuildImageProvider(resolved.image.resolved))
       noteProviderNeed(resolved.image, "Backdrop generation");
+    if (keyframe?.willGenerate && isSupportedBuildImageProvider(resolved.image.resolved))
+      noteProviderNeed(resolved.image, "Keyframe generation");
     if (video?.willGenerate) noteProviderNeed(resolved.video, "Video generation");
     if (music?.willGenerate) noteProviderNeed(resolved.music, "Music generation");
     if (!compositionExists) missing.add("compositions");
@@ -33766,7 +33807,7 @@ async function createBuildPlan(opts) {
       heading: beat.heading,
       durationSec: beat.duration ?? null,
       cues: cue,
-      assets: { narration, backdrop, video, music },
+      assets: { narration, backdrop, keyframe, video, music },
       composition: {
         path: compositionPath,
         exists: compositionExists
@@ -33982,6 +34023,7 @@ function providerResolutionsForPlan(resolved, beats, opts, includeAssets) {
   return [
     includeAssets && !opts.skipNarration && needsProvider("narration") ? resolved.narration : null,
     includeAssets && !opts.skipBackdrop && needsProvider("backdrop") ? resolved.image : null,
+    includeAssets && !opts.skipKeyframe && needsProvider("keyframe") ? resolved.image : null,
     includeAssets && !opts.skipVideo && needsProvider("video") ? resolved.video : null,
     includeAssets && !opts.skipMusic && needsProvider("music") ? resolved.music : null,
     resolved.composer ?? null
@@ -34556,6 +34598,7 @@ async function executeSceneBuild(opts) {
     skipNarration: opts.skipNarration,
     skipBackdrop: opts.skipBackdrop,
     skipVideo: opts.skipVideo,
+    skipKeyframe: opts.skipKeyframe,
     skipMusic: opts.skipMusic,
     ttsProvider: opts.ttsProvider,
     voice: opts.voice,
@@ -34741,6 +34784,7 @@ async function executeSceneBuild(opts) {
         skipNarration: opts.skipNarration ?? false,
         skipBackdrop: opts.skipBackdrop ?? false,
         skipVideo: opts.skipVideo ?? false,
+        skipKeyframe: opts.skipKeyframe ?? false,
         skipMusic: opts.skipMusic ?? false,
         force: opts.force ?? false,
         onProgress,
@@ -35117,10 +35161,11 @@ async function executeSceneBuild(opts) {
   });
 }
 async function buildBeatPrimitives(beat, ctx) {
+  const keyframe = ctx.skipKeyframe ? await skipped("keyframe", beat.id, "--skip-keyframe", ctx) : await dispatchKeyframe(beat, ctx);
   const [narration, backdrop, video, music] = await Promise.all([
     ctx.skipNarration ? skipped("narration", beat.id, "--skip-narration", ctx) : dispatchNarration(beat, ctx),
     ctx.skipBackdrop ? skipped("backdrop", beat.id, "--skip-backdrop", ctx) : dispatchBackdrop(beat, ctx),
-    ctx.skipVideo ? skipped("video", beat.id, "--skip-video", ctx) : dispatchVideo(beat, ctx),
+    ctx.skipVideo ? skipped("video", beat.id, "--skip-video", ctx) : dispatchVideo(beat, ctx, keyframe),
     ctx.skipMusic ? skipped("music", beat.id, "--skip-music", ctx) : dispatchMusic(beat, ctx)
   ]);
   return {
@@ -35164,6 +35209,8 @@ async function buildBeatPrimitives(beat, ctx) {
       videoMetadataPath: video.metadataPath,
       videoFreshness: video.freshness,
       videoSourcePath: video.sourcePath,
+      keyframeStatus: keyframe.status,
+      keyframePath: keyframe.path,
       musicStatus: music.status,
       musicPath: music.path,
       musicJobId: music.job?.id,
@@ -35608,6 +35655,112 @@ async function generateBackdropImage(prompt3, ctx, ratio) {
   });
   return imageBufferFromResult(result);
 }
+async function editKeyframeImage(prompt3, sheetBuffers, ctx, ratio) {
+  loadSceneBuildEnv(ctx.projectDir);
+  const keyInfo = imageProviderKeyInfo(ctx.imageProvider);
+  const apiKey = await getApiKeyFromConfig(keyInfo.configKey, { cwd: ctx.projectDir }) ?? process.env[keyInfo.envVar] ?? "";
+  if (!apiKey) {
+    return {
+      success: false,
+      error: `${keyInfo.envVar} not set \u2014 cannot generate keyframe with ${ctx.imageProvider}`
+    };
+  }
+  if (ctx.imageProvider === "openai") {
+    const provider2 = new OpenAIImageProvider();
+    await provider2.initialize({ apiKey });
+    const result2 = await provider2.editImage(sheetBuffers, prompt3);
+    return imageBufferFromResult(result2);
+  }
+  if (ctx.imageProvider === "gemini") {
+    const provider2 = new GeminiProvider();
+    await provider2.initialize({ apiKey });
+    const result2 = await provider2.editImage(sheetBuffers, prompt3, {
+      model: "flash",
+      aspectRatio: ratio
+    });
+    return imageBufferFromResult(result2);
+  }
+  const provider = new GrokProvider();
+  await provider.initialize({ apiKey });
+  const result = await provider.editImage(sheetBuffers[0], prompt3, { aspectRatio: ratio });
+  return imageBufferFromResult(result);
+}
+async function ensureKeyframe(beat, ctx, keyframeCue, characterRefsRel) {
+  const rel = `assets/keyframe-${beat.id}.png`;
+  const abs = join16(ctx.projectDir, rel);
+  const size = ctx.imageSize ?? "1536x1024";
+  const ratio = imageRatioForSize(size);
+  const metadataOptions = { quality: ctx.imageQuality, size, ratio };
+  const cache = keyframeCacheDescriptor({
+    beatId: beat.id,
+    cue: keyframeCue,
+    provider: ctx.imageProvider,
+    quality: ctx.imageQuality,
+    size,
+    ratio,
+    characters: characterRefsRel
+  });
+  if (existsSync19(abs) && !ctx.force) {
+    const metadata = readAssetMetadata(ctx.projectDir, "keyframe", beat.id);
+    if (!metadata || isFreshCanonicalAsset({
+      projectDir: ctx.projectDir,
+      kind: "keyframe",
+      beatId: beat.id,
+      cue: keyframeCue,
+      provider: ctx.imageProvider,
+      options: metadataOptions,
+      cacheKey: cache.key
+    })) {
+      return { status: "cached", rel };
+    }
+  }
+  const cacheAbs = join16(ctx.projectDir, cache.path);
+  if (existsSync19(cacheAbs) && !ctx.force) {
+    await mkdir10(dirname10(abs), { recursive: true });
+    await copyFile3(cacheAbs, abs);
+    await writeAssetMetadata({
+      projectDir: ctx.projectDir,
+      kind: "keyframe",
+      beatId: beat.id,
+      cue: keyframeCue,
+      provider: ctx.imageProvider,
+      options: metadataOptions,
+      cacheKey: cache.key,
+      canonicalPath: rel,
+      cachePath: cache.path
+    });
+    return { status: "cached", rel };
+  }
+  const sheetBuffers = [];
+  for (const relPath of characterRefsRel) {
+    const sheetAbs = join16(ctx.projectDir, relPath);
+    if (existsSync19(sheetAbs)) sheetBuffers.push(await readFile18(sheetAbs));
+  }
+  const imgCtx = {
+    projectDir: ctx.projectDir,
+    imageProvider: ctx.imageProvider,
+    imageSize: size,
+    imageQuality: ctx.imageQuality
+  };
+  const generated = sheetBuffers.length > 0 ? await editKeyframeImage(keyframeCue, sheetBuffers, imgCtx, ratio) : await generateBackdropImage(keyframeCue, imgCtx, ratio);
+  if (!generated.success) return { status: "failed", error: generated.error };
+  await mkdir10(dirname10(abs), { recursive: true });
+  await writeFile18(abs, generated.buffer);
+  await mkdir10(dirname10(cacheAbs), { recursive: true });
+  await writeFile18(cacheAbs, generated.buffer);
+  await writeAssetMetadata({
+    projectDir: ctx.projectDir,
+    kind: "keyframe",
+    beatId: beat.id,
+    cue: keyframeCue,
+    provider: ctx.imageProvider,
+    options: metadataOptions,
+    cacheKey: cache.key,
+    canonicalPath: rel,
+    cachePath: cache.path
+  });
+  return { status: "generated", rel };
+}
 async function imageBufferFromResult(result) {
   const image = result.images?.[0];
   if (!result.success || !image) {
@@ -35628,10 +35781,25 @@ function imageProviderKeyInfo(provider) {
   if (provider === "grok") return { configKey: "xai", envVar: "XAI_API_KEY" };
   return { configKey: "openai", envVar: "OPENAI_API_KEY" };
 }
-async function dispatchVideo(beat, ctx) {
+async function dispatchKeyframe(beat, ctx) {
+  const keyframeCue = stringOrUndefined4(beat.cues?.keyframe);
+  if (!keyframeCue) return { status: "no-cue" };
+  const characterRefsRel = beatCharacterNames(beat.cues).map((name) => ctx.characterPaths?.get(name)).filter((p) => typeof p === "string");
+  const kf = await ensureKeyframe(beat, ctx, keyframeCue, characterRefsRel);
+  if (kf.status === "failed") {
+    ctx.onProgress({ type: "keyframe-failed", beatId: beat.id, error: kf.error });
+    return { status: "failed", error: kf.error };
+  }
+  ctx.onProgress(
+    kf.status === "cached" ? { type: "keyframe-cached", beatId: beat.id, path: kf.rel } : { type: "keyframe-generated", beatId: beat.id, path: kf.rel, provider: ctx.imageProvider }
+  );
+  return { status: kf.status, path: kf.rel, provider: ctx.imageProvider };
+}
+async function dispatchVideo(beat, ctx, keyframe) {
   const reference = assetReferenceForBeat(ctx.projectDir, "video", beat);
   if (reference) return referencePrimitiveOutcome("video", beat, ctx, reference);
-  const prompt3 = stringOrUndefined4(beat.cues?.video);
+  const keyframeCue = stringOrUndefined4(beat.cues?.keyframe);
+  const prompt3 = stringOrUndefined4(beat.cues?.video) ?? keyframeCue;
   if (!prompt3) return { status: "no-cue" };
   const characterRefsRel = beatCharacterNames(beat.cues).map((name) => ctx.characterPaths?.get(name)).filter((p) => typeof p === "string");
   const characterRefsAbs = characterRefsRel.map((p) => join16(ctx.projectDir, p));
@@ -35642,7 +35810,10 @@ async function dispatchVideo(beat, ctx) {
     cue: prompt3,
     provider: ctx.videoProvider,
     duration: normalizeVideoDuration(beat.duration),
-    characters: characterRefsRel
+    // Keyframe mode keys the clip to the keyframe prompt; refs feed the keyframe,
+    // not the clip directly.
+    characters: keyframeCue ? void 0 : characterRefsRel,
+    keyframe: keyframeCue
   });
   const metadataPath = assetMetadataPath("video", beat.id);
   if (existsSync19(abs) && !ctx.force) {
@@ -35694,6 +35865,13 @@ async function dispatchVideo(beat, ctx) {
       freshness: "fresh"
     };
   }
+  const keyframeImageRel = keyframeCue && (keyframe.status === "generated" || keyframe.status === "cached") ? keyframe.path : void 0;
+  if (keyframeCue && !keyframeImageRel) {
+    const reason = keyframe.status === "failed" ? `keyframe unavailable: ${keyframe.error ?? "generation failed"}` : "keyframe not generated (build without --skip-keyframe first)";
+    ctx.onProgress({ type: "video-skipped", beatId: beat.id, reason });
+    return { status: "skipped" };
+  }
+  const keyframeImageAbs = keyframeImageRel ? join16(ctx.projectDir, keyframeImageRel) : void 0;
   loadSceneBuildEnv(ctx.projectDir);
   const result = await executeVideoGenerate({
     prompt: prompt3,
@@ -35702,7 +35880,10 @@ async function dispatchVideo(beat, ctx) {
     ratio: "16:9",
     output: abs,
     wait: false,
-    refImages: characterRefsAbs.length > 0 ? characterRefsAbs : void 0,
+    // Keyframe mode → single init frame (image-to-video); otherwise character
+    // reference-to-video.
+    image: keyframeImageAbs,
+    refImages: keyframeImageAbs ? void 0 : characterRefsAbs.length > 0 ? characterRefsAbs : void 0,
     apiKey: await apiKeyForVideoProvider(ctx.videoProvider, ctx.projectDir)
   });
   if (!result.success || !result.taskId) {
@@ -35744,7 +35925,7 @@ async function dispatchVideo(beat, ctx) {
     jobType: "generate-video",
     provider: result.provider ?? ctx.videoProvider,
     providerTaskId: result.taskId,
-    providerTaskType: "text2video",
+    providerTaskType: keyframeImageAbs ? "image2video" : "text2video",
     status: "running",
     projectDir: ctx.projectDir,
     workingDirectory: ctx.projectDir,
@@ -36071,6 +36252,7 @@ function estimateActualAssetCost(outcomes, imageQuality) {
   for (const outcome of outcomes) {
     if (outcome.narrationStatus === "generated") cost += 0.05;
     if (outcome.backdropStatus === "generated") cost += backdropCostUsd(imageQuality);
+    if (outcome.keyframeStatus === "generated") cost += backdropCostUsd(imageQuality);
     if (outcome.videoStatus === "generated" || outcome.videoStatus === "pending") {
       cost += outcome.videoProvider === "seedance" || outcome.videoProvider === "fal" ? estimateSeedanceVideoCostUsd({
         durationSec: normalizeVideoDuration(outcome.sceneDurationSec),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vibeframe/cli",
-  "version": "0.113.2",
+  "version": "0.113.4",
   "description": "VibeFrame CLI - agentic video workflow runtime",
   "private": false,
   "license": "MIT",
@@ -83,8 +83,8 @@
     "tsx": "^4.21.0",
     "typescript": "^5.3.3",
     "vitest": "^1.2.2",
-    "@vibeframe/ai-providers": "0.113.2",
-    "@vibeframe/core": "0.113.2"
+    "@vibeframe/ai-providers": "0.113.4",
+    "@vibeframe/core": "0.113.4"
   },
   "engines": {
     "node": ">=20"