@vibeframe/mcp-server 0.105.2 → 0.106.0

package/dist/index.js

@@ -435908,6 +435908,40 @@ var init_whisper = __esm3({
     init_WhisperProvider();
   }
 });
+function isGeminiTextModelAlias(model) {
+  return Object.prototype.hasOwnProperty.call(GEMINI_TEXT_MODEL_ALIASES, model);
+}
+function resolveGeminiTextModel(model) {
+  const trimmed = model?.trim();
+  if (!trimmed)
+    return GEMINI_DEFAULT_TEXT_MODEL;
+  if (isGeminiTextModelAlias(trimmed))
+    return GEMINI_TEXT_MODEL_ALIASES[trimmed];
+  if (trimmed.startsWith("gemini-"))
+    return trimmed;
+  return GEMINI_DEFAULT_TEXT_MODEL;
+}
+var GEMINI_DEFAULT_TEXT_MODEL;
+var GEMINI_AGENT_DEFAULT_TEXT_MODEL;
+var GEMINI_TEXT_MODEL_ALIASES;
+var GEMINI_TEXT_MODEL_HELP;
+var init_gemini_models = __esm3({
+  "../ai-providers/dist/gemini/gemini-models.js"() {
+    "use strict";
+    GEMINI_DEFAULT_TEXT_MODEL = "gemini-3.5-flash";
+    GEMINI_AGENT_DEFAULT_TEXT_MODEL = "gemini-2.5-flash";
+    GEMINI_TEXT_MODEL_ALIASES = {
+      flash: GEMINI_DEFAULT_TEXT_MODEL,
+      latest: GEMINI_DEFAULT_TEXT_MODEL,
+      "flash-3.5": GEMINI_DEFAULT_TEXT_MODEL,
+      "flash-3": "gemini-3-flash-preview",
+      "flash-2.5": "gemini-2.5-flash",
+      pro: "gemini-2.5-pro",
+      "pro-3.1": "gemini-3.1-pro-preview"
+    };
+    GEMINI_TEXT_MODEL_HELP = "flash/latest (Gemini 3.5 Flash), flash-3.5, flash-3, flash-2.5, pro, pro-3.1, or a full gemini-* model ID";
+  }
+});
 function buildMotionSystemPrompt(width, height, fps, duration, durationInFrames, stylePreset) {
   return `You are a world-class broadcast motion graphics designer (like Apple Keynote, Netflix intros, ESPN graphics). Create STUNNING, jaw-dropping overlays that make viewers go "wow".
 
@@ -436060,7 +436094,7 @@ async function generateMotion(api, description, options = {}) {
   const duration = options.duration || 5;
   const durationInFrames = Math.round(duration * fps);
   const stylePreset = options.style || "modern and clean";
-  const modelId = options.model || "gemini-2.5-pro";
+  const modelId = resolveGeminiTextModel(options.model);
   let systemPrompt = buildMotionSystemPrompt(width, height, fps, duration, durationInFrames, stylePreset);
   if (options.videoContext) {
     const sourceLabel = options.sourceType === "image" ? "IMAGE" : "VIDEO";
@@ -436141,7 +436175,7 @@ async function refineMotion(api, existingCode, instructions, options = {}) {
   const fps = options.fps || 30;
   const duration = options.duration || 5;
   const durationInFrames = Math.round(duration * fps);
-  const modelId = options.model || "gemini-2.5-pro";
+  const modelId = resolveGeminiTextModel(options.model);
   const systemPrompt = `You are a world-class broadcast motion graphics designer. Modify the provided Remotion component based on instructions.
 
 CANVAS: ${width}\xD7${height}px | ${fps}fps | ${durationInFrames} frames (${duration}s)
@@ -436237,9 +436271,18 @@ var STYLE_USER_INSTRUCTIONS;
 var init_gemini_motion = __esm3({
   "../ai-providers/dist/gemini/gemini-motion.js"() {
     "use strict";
+    init_gemini_models();
     GEMINI_MOTION_MODELS = {
+      gemini: GEMINI_DEFAULT_TEXT_MODEL,
+      flash: GEMINI_DEFAULT_TEXT_MODEL,
+      latest: GEMINI_DEFAULT_TEXT_MODEL,
+      "flash-3.5": GEMINI_DEFAULT_TEXT_MODEL,
+      "flash-3": "gemini-3-flash-preview",
+      "flash-2.5": "gemini-2.5-flash",
       pro: "gemini-2.5-pro",
-      "3.1-pro": "gemini-3.1-pro-preview"
+      "2.5-pro": "gemini-2.5-pro",
+      "3.1-pro": "gemini-3.1-pro-preview",
+      "pro-3.1": "gemini-3.1-pro-preview"
     };
     STYLE_USER_INSTRUCTIONS = {
       minimal: "CENTER-ALIGNED pure typography. Thin lines + large title + subtitle. NO glass panels, NO bokeh, NO dark bars. Breathing pulse during hold.",
@@ -436358,7 +436401,7 @@ async function analyzeContent(api, content, targetDuration, options) {
         maxOutputTokens: 4096
       }
     };
-    const response = await fetch(`${api.baseUrl}/models/gemini-2.5-flash:generateContent?key=${api.apiKey}`, {
+    const response = await fetch(`${api.baseUrl}/models/${GEMINI_DEFAULT_TEXT_MODEL}:generateContent?key=${api.apiKey}`, {
       method: "POST",
       headers: { "Content-Type": "application/json" },
       body: JSON.stringify(payload)
@@ -436391,6 +436434,7 @@ var init_gemini_storyboard = __esm3({
   "../ai-providers/dist/gemini/gemini-storyboard.js"() {
     "use strict";
     init_storyboard_prompt();
+    init_gemini_models();
   }
 });
 async function fetchJson(label, url, init) {
@@ -436427,6 +436471,7 @@ var init_GeminiProvider = __esm3({
   "../ai-providers/dist/gemini/GeminiProvider.js"() {
     "use strict";
     init_gemini_motion();
+    init_gemini_models();
     init_gemini_storyboard();
     init_http();
     MODEL_MAP = {
@@ -436935,7 +436980,7 @@ var init_GeminiProvider = __esm3({
             error: "Google API key not configured"
           };
         }
-        const modelId = options.model || "gemini-3-flash-preview";
+        const modelId = resolveGeminiTextModel(options.model);
         try {
           let videoPart;
           if (typeof videoData === "string") {
@@ -437042,7 +437087,7 @@ var init_GeminiProvider = __esm3({
             error: "Google API key not configured"
           };
         }
-        const modelId = options.model || "gemini-3-flash-preview";
+        const modelId = resolveGeminiTextModel(options.model);
         try {
           const buffers = Array.isArray(imageData) ? imageData : [imageData];
           const imageParts = buffers.map((buf) => ({
@@ -437144,7 +437189,7 @@ Example response:
 [{"type":"trim","description":"Trim intro to 3 seconds","clipIds":["clip-1"],"params":{"newDuration":3},"confidence":0.9}]
 
 Respond with ONLY the JSON array, no other text.`;
-          const response = await fetch(`${this.baseUrl}/models/gemini-2.5-flash:generateContent?key=${this.apiKey}`, {
+          const response = await fetch(`${this.baseUrl}/models/${GEMINI_DEFAULT_TEXT_MODEL}:generateContent?key=${this.apiKey}`, {
             method: "POST",
             headers: {
               "Content-Type": "application/json"
@@ -437264,6 +437309,7 @@ var init_gemini = __esm3({
     "use strict";
     init_GeminiProvider();
     init_gemini_motion();
+    init_gemini_models();
     init_define_provider();
     defineProvider({
       id: "gemini",
@@ -446526,6 +446572,10 @@ __export3(dist_exports22, {
   ClaudeProvider: () => ClaudeProvider,
   ElevenLabsProvider: () => ElevenLabsProvider,
   FalProvider: () => FalProvider,
+  GEMINI_AGENT_DEFAULT_TEXT_MODEL: () => GEMINI_AGENT_DEFAULT_TEXT_MODEL,
+  GEMINI_DEFAULT_TEXT_MODEL: () => GEMINI_DEFAULT_TEXT_MODEL,
+  GEMINI_TEXT_MODEL_ALIASES: () => GEMINI_TEXT_MODEL_ALIASES,
+  GEMINI_TEXT_MODEL_HELP: () => GEMINI_TEXT_MODEL_HELP,
   GeminiProvider: () => GeminiProvider,
   GrokProvider: () => GrokProvider,
   KNOWN_VOICES: () => KNOWN_VOICES,
@@ -446556,6 +446606,7 @@ __export3(dist_exports22, {
   getProvidersFor: () => getProvidersFor,
   getSetupProviders: () => getSetupProviders,
   grokProvider: () => grokProvider,
+  isGeminiTextModelAlias: () => isGeminiTextModelAlias,
   klingProvider: () => klingProvider,
   kokoroProvider: () => kokoroProvider,
   ollamaProvider: () => ollamaProvider,
@@ -446563,6 +446614,7 @@ __export3(dist_exports22, {
   openaiProvider: () => openaiProvider,
   providerRegistry: () => providerRegistry,
   replicateProvider: () => replicateProvider,
+  resolveGeminiTextModel: () => resolveGeminiTextModel,
   resolveVoiceId: () => resolveVoiceId,
   runwayProvider: () => runwayProvider,
   whisperProvider: () => whisperProvider
@@ -449708,6 +449760,7 @@ var CONFIG_PROVIDER_BY_COMPOSER;
 var init_compose_scenes_skills = __esm3({
   "src/commands/_shared/compose-scenes-skills.ts"() {
     "use strict";
+    init_dist22();
     init_config();
     init_bundle();
     init_scene_lint();
@@ -449729,12 +449782,12 @@ var init_compose_scenes_skills = __esm3({
         medium: { model: "gpt-5", maxTokens: 6e3, costPerMTokIn: 1.25, costPerMTokOut: 10 },
         high: { model: "gpt-5", maxTokens: 8e3, costPerMTokIn: 1.25, costPerMTokOut: 10 }
       },
-      // Google Gemini 2.5 Pro — ~2.6× cheaper than Claude per spike, ~20 s/beat.
-      // Strong instruction-following on the Hyperframes skill bundle.
+      // Google Gemini 3.5 Flash — current one-shot Gemini default for
+      // analysis/composition. Keep agent loops on their explicit 2.5 Flash default.
       gemini: {
-        low: { model: "gemini-2.5-pro", maxTokens: 4e3, costPerMTokIn: 1.25, costPerMTokOut: 5 },
-        medium: { model: "gemini-2.5-pro", maxTokens: 6e3, costPerMTokIn: 1.25, costPerMTokOut: 5 },
-        high: { model: "gemini-2.5-pro", maxTokens: 8e3, costPerMTokIn: 1.25, costPerMTokOut: 5 }
+        low: { model: GEMINI_DEFAULT_TEXT_MODEL, maxTokens: 4e3, costPerMTokIn: 1.5, costPerMTokOut: 9 },
+        medium: { model: GEMINI_DEFAULT_TEXT_MODEL, maxTokens: 6e3, costPerMTokIn: 1.5, costPerMTokOut: 9 },
+        high: { model: GEMINI_DEFAULT_TEXT_MODEL, maxTokens: 8e3, costPerMTokIn: 1.5, costPerMTokOut: 9 }
       }
     };
     ComposeBeatError = class extends Error {
@@ -456309,12 +456362,7 @@ async function detectSilencePeriodsWithGemini(videoPath, minDuration, options) {
   const gemini = new GeminiProvider();
   await gemini.initialize({ apiKey: geminiApiKey });
   const videoBuffer = await readFile18(videoPath);
-  const modelMap = {
-    flash: "gemini-3-flash-preview",
-    "flash-2.5": "gemini-2.5-flash",
-    pro: "gemini-2.5-pro"
-  };
-  const modelId = options.model ? modelMap[options.model] || modelMap.flash : void 0;
+  const modelId = options.model ? resolveGeminiTextModel(options.model) : void 0;
   const prompt3 = `Analyze this video and identify all silent or dead segments where there is NO meaningful content.
 
 Detect these as silent/dead segments:
@@ -459710,7 +459758,7 @@ Use this image analysis to inform the color palette, typography placement, and o
 6. Recommended animation timing and entrance/exit style
 7. Any moments where overlays should pause, fade, or stay minimal`;
       const analysisResult = await gemini.analyzeVideo(videoBuffer, analysisPrompt, {
-        model: "gemini-2.5-flash",
+        model: GEMINI_DEFAULT_TEXT_MODEL,
         fps: 1,
         lowResolution: true
       });
@@ -459879,7 +459927,7 @@ function registerMotionCommand(aiCommand) {
     "auto"
   ).option("--understanding-prompt <text>", "Custom prompt for --video understanding").option("--from-tsx <path>", "Refine an existing TSX file instead of generating from scratch").option(
     "-m, --model <alias>",
-    "LLM model: sonnet (default), opus, gemini, gemini-3.1-pro",
+    "LLM model: sonnet (default), opus, gemini, gemini-2.5-pro, gemini-3.1-pro",
     "sonnet"
   ).option("--dry-run", "Preview parameters without executing").action(async (description, options) => {
     const startedAt = Date.now();
@@ -460023,7 +460071,8 @@ var init_ai_motion = __esm3({
       sonnet: { provider: "claude", modelId: "claude-sonnet-4-6" },
       opus: { provider: "claude", modelId: "claude-opus-4-7" },
       "opus-4-6": { provider: "claude", modelId: "claude-opus-4-6" },
-      gemini: { provider: "gemini", modelId: "gemini-2.5-pro" },
+      gemini: { provider: "gemini", modelId: GEMINI_DEFAULT_TEXT_MODEL },
+      "gemini-2.5-pro": { provider: "gemini", modelId: "gemini-2.5-pro" },
       "gemini-3.1-pro": { provider: "gemini", modelId: "gemini-3.1-pro-preview" }
     };
   }
@@ -460584,7 +460633,7 @@ async function executeMotionOverlay(options) {
   };
 }
 function registerMotionOverlayCommand(parent) {
-  parent.command("motion-overlay").description("Apply designed motion graphics overlays to an existing video").argument("<video>", "Video file path").argument("[description]", "Motion overlay description (omit when using --asset)").option("--asset <path>", "User-provided .json/.lottie animation to overlay").option("-o, --output <path>", "Output video file path").option("-d, --duration <sec>", "Overlay/render duration in seconds").option("--start <sec>", "Overlay start time in seconds", "0").option("--style <style>", "Style preset for generated overlays: minimal, corporate, playful, cinematic").option("-m, --model <alias>", "LLM model for generated overlays: sonnet, opus, gemini, gemini-3.1-pro", "sonnet").option("--understand <mode>", "Analyze video before generated overlay: auto, off, required", "auto").option("--understanding-prompt <text>", "Custom prompt for video understanding").option("--position <position>", "Lottie position: full, center, top-left, top-right, bottom-left, bottom-right", "full").option("--scale <number>", "Lottie overlay scale (0.01-2)").option("--opacity <number>", "Lottie overlay opacity (0-1)", "1").option("--loop", "Loop Lottie overlay", true).option("--no-loop", "Do not loop Lottie overlay").option("--dry-run", "Preview parameters without executing").action(async (videoPath, description, options) => {
+  parent.command("motion-overlay").description("Apply designed motion graphics overlays to an existing video").argument("<video>", "Video file path").argument("[description]", "Motion overlay description (omit when using --asset)").option("--asset <path>", "User-provided .json/.lottie animation to overlay").option("-o, --output <path>", "Output video file path").option("-d, --duration <sec>", "Overlay/render duration in seconds").option("--start <sec>", "Overlay start time in seconds", "0").option("--style <style>", "Style preset for generated overlays: minimal, corporate, playful, cinematic").option("-m, --model <alias>", "LLM model for generated overlays: sonnet, opus, gemini, gemini-2.5-pro, gemini-3.1-pro", "sonnet").option("--understand <mode>", "Analyze video before generated overlay: auto, off, required", "auto").option("--understanding-prompt <text>", "Custom prompt for video understanding").option("--position <position>", "Lottie position: full, center, top-left, top-right, bottom-left, bottom-right", "full").option("--scale <number>", "Lottie overlay scale (0.01-2)").option("--opacity <number>", "Lottie overlay opacity (0-1)", "1").option("--loop", "Loop Lottie overlay", true).option("--no-loop", "Do not loop Lottie overlay").option("--dry-run", "Preview parameters without executing").action(async (videoPath, description, options) => {
     const startedAt = Date.now();
     try {
       if (options.output) validateOutputPath(options.output);
@@ -461864,13 +461913,7 @@ async function executeThumbnailBestFrame(options) {
     await gemini.initialize({ apiKey: googleKey });
     const videoData = await readFile25(videoPath);
     const analysisPrompt = prompt3 || 'Analyze this video and find the single best frame for a thumbnail. Look for frames that are visually striking, well-composed, emotionally engaging, and representative of the video content. Avoid blurry frames, transitions, or dark scenes. Return ONLY a JSON object: {"timestamp": <seconds as number>, "reason": "<brief explanation>"}';
-    const modelMap = {
-      flash: "gemini-3-flash-preview",
-      latest: "gemini-2.5-flash",
-      "flash-2.5": "gemini-2.5-flash",
-      pro: "gemini-2.5-pro"
-    };
-    const modelId = modelMap[model] || "gemini-3-flash-preview";
+    const modelId = resolveGeminiTextModel(model);
     const result = await gemini.analyzeVideo(videoData, analysisPrompt, {
       model: modelId,
       fps: 1
@@ -461925,12 +461968,7 @@ async function executeGeminiVideo(options) {
       return { success: false, error: "Google API key required. Run 'vibe setup' or set GOOGLE_API_KEY in .env" };
     }
     const isYouTube = options.source.includes("youtube.com") || options.source.includes("youtu.be");
-    const modelMap = {
-      flash: "gemini-3-flash-preview",
-      "flash-2.5": "gemini-2.5-flash",
-      pro: "gemini-2.5-pro"
-    };
-    const modelId = modelMap[options.model || "flash"] || modelMap.flash;
+    const modelId = resolveGeminiTextModel(options.model);
     let videoData;
     if (isYouTube) {
       videoData = options.source;
@@ -461991,12 +462029,7 @@ async function executeAnalyze(options) {
         error: "Cannot detect source type. Supported: images (.png/.jpg/.webp/.gif), videos (.mp4/.mov/.webm), YouTube URLs, image URLs."
       };
     }
-    const modelMap = {
-      flash: "gemini-3-flash-preview",
-      "flash-2.5": "gemini-2.5-flash",
-      pro: "gemini-2.5-pro"
-    };
-    const modelId = modelMap[options.model || "flash"] || modelMap.flash;
+    const modelId = resolveGeminiTextModel(options.model);
     const gemini = new GeminiProvider();
     await gemini.initialize({ apiKey });
     if (isImage) {
@@ -462076,7 +462109,7 @@ async function executeAnalyze(options) {
   }
 }
 function registerAnalyzeCommands(aiCommand) {
-  aiCommand.command("gemini-video").description("Analyze video using Gemini (summarize, Q&A, extract info)").argument("<source>", "Video file path or YouTube URL").argument("<prompt>", "Analysis prompt (e.g., 'Summarize this video')").option("-k, --api-key <key>", "Google API key (or set GOOGLE_API_KEY env)").option("-m, --model <model>", "Model: flash (default), flash-2.5, pro", "flash").option("--fps <number>", "Frames per second (default: 1, higher for action)").option("--start <seconds>", "Start offset in seconds (for clipping)").option("--end <seconds>", "End offset in seconds (for clipping)").option("--low-res", "Use low resolution mode (fewer tokens, longer videos)").option("-v, --verbose", "Show token usage").action(async (source3, prompt3, options) => {
+  aiCommand.command("gemini-video").description("Analyze video using Gemini (summarize, Q&A, extract info)").argument("<source>", "Video file path or YouTube URL").argument("<prompt>", "Analysis prompt (e.g., 'Summarize this video')").option("-k, --api-key <key>", "Google API key (or set GOOGLE_API_KEY env)").option("-m, --model <model>", `Model: ${GEMINI_TEXT_MODEL_HELP}`, "flash").option("--fps <number>", "Frames per second (default: 1, higher for action)").option("--start <seconds>", "Start offset in seconds (for clipping)").option("--end <seconds>", "End offset in seconds (for clipping)").option("--low-res", "Use low resolution mode (fewer tokens, longer videos)").option("-v, --verbose", "Show token usage").action(async (source3, prompt3, options) => {
     try {
       if (options.apiKey) {
         process.env.GOOGLE_API_KEY = options.apiKey;
@@ -462119,7 +462152,7 @@ function registerAnalyzeCommands(aiCommand) {
       exitWithError(apiError(`Video analysis failed: ${error instanceof Error ? error.message : String(error)}`, true));
     }
   });
-  aiCommand.command("analyze").description("Analyze any media: images, videos, or YouTube URLs using Gemini").argument("<source>", "Image/video file path, image URL, or YouTube URL").argument("<prompt>", "Analysis prompt (e.g., 'Describe this image', 'Summarize this video')").option("-k, --api-key <key>", "Google API key (or set GOOGLE_API_KEY env)").option("-m, --model <model>", "Model: flash (default), flash-2.5, pro", "flash").option("--fps <number>", "Frames per second for video (default: 1)").option("--start <seconds>", "Start offset in seconds (video only)").option("--end <seconds>", "End offset in seconds (video only)").option("--low-res", "Use low resolution mode (fewer tokens)").option("-v, --verbose", "Show token usage").action(async (source3, prompt3, options) => {
+  aiCommand.command("analyze").description("Analyze any media: images, videos, or YouTube URLs using Gemini").argument("<source>", "Image/video file path, image URL, or YouTube URL").argument("<prompt>", "Analysis prompt (e.g., 'Describe this image', 'Summarize this video')").option("-k, --api-key <key>", "Google API key (or set GOOGLE_API_KEY env)").option("-m, --model <model>", `Model: ${GEMINI_TEXT_MODEL_HELP}`, "flash").option("--fps <number>", "Frames per second for video (default: 1)").option("--start <seconds>", "Start offset in seconds (video only)").option("--end <seconds>", "End offset in seconds (video only)").option("--low-res", "Use low resolution mode (fewer tokens)").option("-v, --verbose", "Show token usage").action(async (source3, prompt3, options) => {
     try {
       if (options.apiKey) {
         process.env.GOOGLE_API_KEY = options.apiKey;
@@ -462235,12 +462268,7 @@ ${content}`;
 
 Project context for beat-aware review:
 ${projectContext}` : "";
-  const modelMap = {
-    flash: "gemini-3-flash-preview",
-    "flash-2.5": "gemini-2.5-flash",
-    pro: "gemini-2.5-pro"
-  };
-  const modelId = modelMap[model] || modelMap.flash;
+  const modelId = resolveGeminiTextModel(model);
   const reviewPrompt = `You are a professional video editor reviewing this video for quality. Analyze the video and return a JSON review with the following structure. Return ONLY valid JSON, no extra text.
 
 {
@@ -462347,7 +462375,7 @@ Score each category 1-10. Prefer beatIssues when you can map a problem to a stor
   return result;
 }
 function registerReviewCommand(aiCommand) {
-  aiCommand.command("review", { hidden: true }).description("Review video quality using Gemini AI and optionally auto-fix issues").argument("<source>", "Video file path").option("--storyboard <path>", "Storyboard JSON file for context").option("--auto-apply", "Automatically apply fixable corrections").option("--verify", "Run verification pass after applying fixes").option("-m, --model <model>", "Gemini model: flash (default), flash-2.5, pro", "flash").option("-o, --output <path>", "Output video file path (for auto-apply)").option("--dry-run", "Preview parameters without executing").action(async (videoPath, options) => {
+  aiCommand.command("review", { hidden: true }).description("Review video quality using Gemini AI and optionally auto-fix issues").argument("<source>", "Video file path").option("--storyboard <path>", "Storyboard JSON file for context").option("--auto-apply", "Automatically apply fixable corrections").option("--verify", "Run verification pass after applying fixes").option("-m, --model <model>", `Gemini model: ${GEMINI_TEXT_MODEL_HELP}`, "flash").option("-o, --output <path>", "Output video file path (for auto-apply)").option("--dry-run", "Preview parameters without executing").action(async (videoPath, options) => {
     const startedAt = Date.now();
     try {
       if (options.output) {
@@ -463204,7 +463232,7 @@ var init_speech = __esm3({
   }
 });
 function registerThumbnailCommand(parent) {
-  parent.command("thumbnail").description("Generate video thumbnail (DALL-E) or extract best frame from video (Gemini)").argument("[description]", "Thumbnail description (for DALL-E generation)").option("-k, --api-key <key>", "API key (OpenAI for generation, Google for best-frame)").option("-o, --output <path>", "Output file path").option("--style <style>", "Platform style: youtube, instagram, tiktok, twitter").option("--best-frame <video>", "Extract best thumbnail frame from video using Gemini AI").option("--prompt <prompt>", "Custom prompt for best-frame analysis").option("--model <model>", "Gemini model: flash, latest, pro (default: flash)", "flash").action(async (description, options) => {
+  parent.command("thumbnail").description("Generate video thumbnail (DALL-E) or extract best frame from video (Gemini)").argument("[description]", "Thumbnail description (for DALL-E generation)").option("-k, --api-key <key>", "API key (OpenAI for generation, Google for best-frame)").option("-o, --output <path>", "Output file path").option("--style <style>", "Platform style: youtube, instagram, tiktok, twitter").option("--best-frame <video>", "Extract best thumbnail frame from video using Gemini AI").option("--prompt <prompt>", "Custom prompt for best-frame analysis").option("--model <model>", `Gemini model: ${GEMINI_TEXT_MODEL_HELP}`, "flash").action(async (description, options) => {
     const startedAt = Date.now();
     try {
       if (description) rejectControlChars(description);
@@ -465404,7 +465432,7 @@ function registerVideoCommand(parent) {
     "--seedance-model <model>",
     "Seedance variant: quality or fast (fal.ai only)",
     "quality"
-  ).option("--negative <prompt>", "Negative prompt - what to avoid (Kling/Veo)").option("--resolution <res>", "Video resolution: 720p, 1080p, 4k (Veo only)").option("--last-frame <path>", "Last frame image for frame interpolation (Veo only)").option(
+  ).option("--negative <prompt>", "Negative prompt - what to avoid (Kling/Veo)").option("--resolution <res>", "Video resolution: 480p, 720p, 1080p, or 4k depending on provider").option("--last-frame <path>", "Last frame image for frame interpolation (Veo) or Seedance end frame").option("--end-image <path>", "Ending frame image for Seedance image-to-video").option(
     "--ref-images <paths...>",
     "Reference images for Seedance reference-to-video or Veo character consistency"
   ).option("--ref-videos <paths...>", "Reference videos for Seedance reference-to-video").option("--ref-audio <paths...>", "Reference audio for Seedance reference-to-video").option("--no-generate-audio", "Disable native audio when the provider supports it").option("--person <mode>", "Person generation: allow_all, allow_adult (Veo only)").option("--veo-model <model>", "Veo model: 3.0, 3.1, 3.1-fast (default: 3.1-fast)", "3.1-fast").option(
@@ -465865,6 +465893,30 @@ Examples:
             exitWithError(apiError(message, true));
           }
         }
+        let seedanceEndImage;
+        const seedanceEndImagePath = options.endImage ?? options.lastFrame;
+        if (seedanceEndImagePath && seedanceReferences.length === 0) {
+          try {
+            const absEndImagePath = resolve50(process.cwd(), seedanceEndImagePath);
+            const endImageBuffer = await readFile32(absEndImagePath);
+            const ext = seedanceEndImagePath.toLowerCase().split(".").pop();
+            const mimeType = ext === "jpg" || ext === "jpeg" ? "image/jpeg" : `image/${ext || "png"}`;
+            const uploadHost = await resolveUploadHost();
+            spinner2.text = `Uploading end image via ${uploadHost.provider} for Seedance...`;
+            const upload = await uploadHost.uploadImage(endImageBuffer, {
+              filename: seedanceEndImagePath,
+              mimeType
+            });
+            seedanceEndImage = upload.url;
+          } catch (err) {
+            spinner2.fail("End image upload failed");
+            const message = err instanceof Error ? err.message : String(err);
+            if (message.includes("IMGBB_API_KEY")) {
+              exitWithError(authError("IMGBB_API_KEY", "ImgBB"));
+            }
+            exitWithError(apiError(message, true));
+          }
+        }
         spinner2.text = "Generating video with fal.ai Seedance 2.0 (this may take 1-3 minutes)...";
         const seedanceModel = String(options.seedanceModel ?? "quality").toLowerCase();
         const falModel = seedanceModel === "fast" || seedanceModel === "seedance-2.0-fast" ? "seedance-2.0-fast" : "seedance-2.0";
@@ -465877,7 +465929,8 @@ Examples:
           negativePrompt: options.negative,
           model: falModel,
           resolution: options.resolution,
-          generateAudio: options.generateAudio
+          generateAudio: options.generateAudio,
+          lastFrame: seedanceEndImage
         });
         finalResult = result;
       }
@@ -471712,6 +471765,14 @@ var SCENE_PRESETS = [
   "kinetic-type",
   "product-shot"
 ];
+var LOTTIE_POSITIONS = [
+  "full",
+  "center",
+  "top-left",
+  "top-right",
+  "bottom-left",
+  "bottom-right"
+];
 var GSAP_CDN = "https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js";
 function esc(text) {
   return text.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;").replace(/"/g, "&quot;").replace(/'/g, "&#39;");
@@ -471992,6 +472053,56 @@ function buildPreset(input3) {
     }
   }
 }
+var DOTLOTTIE_WC_VERSION = "0.9.12";
+var DOTLOTTIE_WEB_VERSION = "0.71.0";
+var DOTLOTTIE_WC_CDN = `https://cdn.jsdelivr.net/npm/@lottiefiles/dotlottie-wc@${DOTLOTTIE_WC_VERSION}/dist/index.js`;
+var DOTLOTTIE_WASM_CDN = `https://cdn.jsdelivr.net/npm/@lottiefiles/dotlottie-web@${DOTLOTTIE_WEB_VERSION}/dist/dotlottie-player.wasm`;
+function lottieOverlayStyle(input3) {
+  const pos = input3.position ?? "full";
+  const rawScale = input3.scale ?? (pos === "full" ? 1 : 0.25);
+  const scale = Math.max(0.01, Math.min(2, rawScale));
+  const opacity = Math.max(0, Math.min(1, input3.opacity ?? 1));
+  const base = [
+    "position:absolute",
+    "pointer-events:none",
+    `opacity:${opacity}`
+  ];
+  if (pos === "full") {
+    base.push("inset:0", "width:100%", "height:100%");
+  } else {
+    const pct = scale * 100;
+    base.push(`width:${pct}%`, `height:${pct}%`);
+    switch (pos) {
+      case "center":
+        base.push("top:50%", "left:50%", "transform:translate(-50%,-50%)");
+        break;
+      case "top-left":
+        base.push("top:4%", "left:4%");
+        break;
+      case "top-right":
+        base.push("top:4%", "right:4%");
+        break;
+      case "bottom-left":
+        base.push("bottom:4%", "left:4%");
+        break;
+      case "bottom-right":
+        base.push("bottom:4%", "right:4%");
+        break;
+    }
+  }
+  return base.join(";");
+}
+function buildLottieOverlay(input3, sceneId) {
+  const loop = input3.loop ?? true ? " loop" : "";
+  const style = lottieOverlayStyle(input3);
+  const overlayId = `lottie-overlay-${sceneId}`;
+  const markup = `<dotlottie-wc id="${overlayId}" src="${esc(input3.src)}" autoplay${loop} style="${style}"></dotlottie-wc>`;
+  const script = `<script type="module">
+      import { setWasmUrl } from "${DOTLOTTIE_WC_CDN}";
+      setWasmUrl("${DOTLOTTIE_WASM_CDN}");
+    </script>`;
+  return { markup, script };
+}
 function emitSceneHtml(input3) {
   if (input3.duration <= 0) {
     throw new Error(`Scene duration must be > 0, got ${input3.duration}`);
@@ -472012,15 +472123,20 @@ function emitSceneHtml(input3) {
       data-volume="1"
     ></audio>
 ` : "";
+  const lottieLayer = input3.lottie ? buildLottieOverlay(input3.lottie, id) : null;
+  const lottieMarkup = lottieLayer ? `
+    ${lottieLayer.markup}` : "";
+  const lottieScript = lottieLayer ? `
+    ${lottieLayer.script}` : "";
   return `<template id="scene-${id}-template">
   <div data-composition-id="${id}" data-start="0" data-duration="${dur}" data-width="${input3.width}" data-height="${input3.height}">
     <style>
       ${parts.css}
     </style>
 
-    ${parts.body}
+    ${parts.body}${lottieMarkup}
 ${audioBlock}
-    <script src="${GSAP_CDN}"></script>
+    <script src="${GSAP_CDN}"></script>${lottieScript}
     <script>
       window.__timelines = window.__timelines || {};
       const tl = gsap.timeline({ paused: true });
@@ -472435,9 +472551,40 @@ sceneCommand.command("add").description("Add a new scene to a project: AI narrat
 ).option("--no-image", "Skip image generation even when --visuals is provided").option(
   "--no-transcribe",
   "Skip Whisper word-level transcribe step (no transcript-<id>.json emitted)"
-).option("--transcribe-language <code>", "BCP-47 language code passed to Whisper (e.g. en, ko)").option("--force", "Overwrite an existing compositions/scene-<id>.html").option("--dry-run", "Preview parameters without writing files or calling APIs").action(async (name, options) => {
+).option("--transcribe-language <code>", "BCP-47 language code passed to Whisper (e.g. en, ko)").option("--force", "Overwrite an existing compositions/scene-<id>.html").option("--lottie <path>", "Lottie animation file (.json/.lottie) to overlay on the scene").option(
+  "--lottie-position <position>",
+  `Lottie position: ${LOTTIE_POSITIONS.join(", ")}`,
+  "full"
+).option("--lottie-scale <number>", "Lottie overlay scale (0.01-2)").option("--lottie-opacity <number>", "Lottie overlay opacity (0-1)", "1").option("--lottie-no-loop", "Do not loop the Lottie animation").option("--dry-run", "Preview parameters without writing files or calling APIs").action(async (name, options) => {
   const startedAt = Date.now();
   if (options.style) options.style = validatePreset(options.style);
+  if (options.lottiePosition) {
+    if (!LOTTIE_POSITIONS.includes(options.lottiePosition)) {
+      exitWithError(
+        usageError(
+          `Invalid --lottie-position "${options.lottiePosition}". Valid: ${LOTTIE_POSITIONS.join(", ")}`
+        )
+      );
+    }
+  }
+  let lottieScale;
+  if (options.lottieScale !== void 0) {
+    lottieScale = Number(options.lottieScale);
+    if (!Number.isFinite(lottieScale) || lottieScale < 0.01 || lottieScale > 2) {
+      exitWithError(
+        usageError(`Invalid --lottie-scale "${options.lottieScale}". Must be between 0.01 and 2.`)
+      );
+    }
+  }
+  let lottieOpacity = 1;
+  if (options.lottieOpacity !== void 0) {
+    lottieOpacity = Number(options.lottieOpacity);
+    if (!Number.isFinite(lottieOpacity) || lottieOpacity < 0 || lottieOpacity > 1) {
+      exitWithError(
+        usageError(`Invalid --lottie-opacity "${options.lottieOpacity}". Must be between 0 and 1.`)
+      );
+    }
+  }
   if (options.duration !== void 0) options.duration = validateDuration(options.duration);
   let tts;
   try {
@@ -472466,7 +472613,11 @@ sceneCommand.command("add").description("Add a new scene to a project: AI narrat
           imageProvider: options.imageProvider,
           tts,
           audio: options.audio,
-          // commander sets `audio: false` when --no-audio is passed
+          lottie: options.lottie ?? null,
+          lottiePosition: options.lottiePosition,
+          lottieScale: lottieScale ?? null,
+          lottieOpacity: options.lottieOpacity !== void 0 ? lottieOpacity : null,
+          lottieLoop: !options.lottieNoLoop,
           image: options.image
         }
       }
@@ -472494,6 +472645,13 @@ sceneCommand.command("add").description("Add a new scene to a project: AI narrat
       skipTranscribe: options.transcribe === false,
       transcribeLanguage: options.transcribeLanguage,
       force: !!options.force,
+      lottie: options.lottie ? {
+        src: options.lottie,
+        position: options.lottiePosition ?? "full",
+        scale: lottieScale,
+        opacity: options.lottieOpacity !== void 0 ? lottieOpacity : 1,
+        loop: !options.lottieNoLoop
+      } : void 0,
       onProgress: (msg) => {
         if (spinner2) spinner2.text = msg;
       }
@@ -472769,6 +472927,30 @@ async function executeSceneAdd(opts) {
       await writeFile10(imageAbsPath, buffer);
     }
   }
+  let lottieRelPath;
+  let lottieAbsPath;
+  let lottieInput;
+  if (opts.lottie) {
+    const sourceAbs = resolve122(opts.lottie.src);
+    if (!await pathExists2(sourceAbs)) {
+      return errResult(`Lottie file not found: ${sourceAbs}`);
+    }
+    const ext = (sourceAbs.match(/\.([a-z0-9]+)$/i)?.[1] ?? "json").toLowerCase();
+    if (ext !== "json" && ext !== "lottie") {
+      return errResult(`Unsupported Lottie file extension: .${ext}. Use .json or .lottie.`);
+    }
+    lottieRelPath = `assets/lottie-${id}.${ext}`;
+    lottieAbsPath = resolve122(projectDir, lottieRelPath);
+    await mkdir5(dirname33(lottieAbsPath), { recursive: true });
+    await copyFile6(sourceAbs, lottieAbsPath);
+    lottieInput = {
+      src: lottieRelPath,
+      position: opts.lottie.position,
+      scale: opts.lottie.scale,
+      opacity: opts.lottie.opacity,
+      loop: opts.lottie.loop
+    };
+  }
   const cfg = await loadVibeProjectConfig(projectDir);
   const fallback2 = cfg?.defaultSceneDuration ?? 5;
   const NARRATION_TAIL_BUFFER = 0.5;
@@ -472797,7 +472979,8 @@ async function executeSceneAdd(opts) {
     kicker: opts.kicker,
     imagePath: imageRelPath,
     audioPath: audioRelPath,
-    transcript: transcriptWords
+    transcript: transcriptWords,
+    lottie: lottieInput
   });
   await mkdir5(dirname33(scenePath), { recursive: true });
   await writeFile10(scenePath, sceneHtml, "utf-8");
@@ -472818,6 +473001,7 @@ async function executeSceneAdd(opts) {
     rootPath: relative5(process.cwd(), rootPath) || rootPath,
     audioPath: audioAbsPath ? relative5(process.cwd(), audioAbsPath) || audioAbsPath : void 0,
     imagePath: imageAbsPath ? relative5(process.cwd(), imageAbsPath) || imageAbsPath : void 0,
+    lottiePath: lottieAbsPath ? relative5(process.cwd(), lottieAbsPath) || lottieAbsPath : void 0,
     transcriptPath: transcriptAbsPath ? relative5(process.cwd(), transcriptAbsPath) || transcriptAbsPath : void 0,
     transcriptWordCount
   };
@@ -474070,7 +474254,7 @@ var editMotionOverlayTool = defineTool({
     duration: external_exports.number().optional().describe("Overlay/render duration in seconds"),
     start: external_exports.number().optional().describe("Overlay start time in seconds"),
     style: external_exports.string().optional().describe("Style preset for generated overlays"),
-    model: external_exports.enum(["sonnet", "opus", "gemini", "gemini-3.1-pro"]).optional().describe("LLM model for generated overlays"),
+    model: external_exports.enum(["sonnet", "opus", "gemini", "gemini-2.5-pro", "gemini-3.1-pro"]).optional().describe("LLM model for generated overlays"),
     understand: external_exports.enum(["auto", "off", "required"]).optional().describe("Analyze video before generated overlay (default: auto)"),
     understandingPrompt: external_exports.string().optional().describe("Custom prompt for video understanding"),
     position: external_exports.enum(["full", "center", "top-left", "top-right", "bottom-left", "bottom-right"]).optional().describe("Lottie overlay position"),
@@ -475785,7 +475969,7 @@ var inspectRenderTool = defineTool({
     outputPath: external_exports.string().optional().describe("Optional review report path. Defaults to <project>/review-report.json."),
     report: external_exports.boolean().optional().describe("Write review-report.json. Default true."),
     ai: external_exports.boolean().optional().describe("Also run Gemini video review and merge findings into review-report.json. Default false."),
-    model: external_exports.enum(["flash", "flash-2.5", "pro"]).optional().describe("Gemini model variant for ai review. Default flash."),
+    model: external_exports.enum(["flash", "latest", "flash-3", "flash-2.5", "pro", "pro-3.1"]).optional().describe("Gemini model variant for ai review. Default flash."),
     dryRun: external_exports.boolean().optional().describe("Preview resolved inputs without probing video or calling Gemini.")
   }),
   async execute(args, ctx) {
@@ -475837,7 +476021,7 @@ var analyzeMediaTool = defineTool({
   schema: external_exports.object({
     source: external_exports.string().describe("Path to image/video or YouTube URL"),
     prompt: external_exports.string().describe("Analysis prompt (e.g., 'Describe the scene', 'Count people')"),
-    model: external_exports.enum(["flash", "flash-2.5", "pro"]).optional().describe("Gemini model variant (default: flash)"),
+    model: external_exports.enum(["flash", "latest", "flash-3", "flash-2.5", "pro", "pro-3.1"]).optional().describe("Gemini model variant (default: flash)"),
     fps: external_exports.number().optional().describe("Frames per second for video sampling (default: 1)"),
     start: external_exports.number().optional().describe("Start time in seconds for video analysis"),
     end: external_exports.number().optional().describe("End time in seconds for video analysis"),
@@ -475865,7 +476049,7 @@ var analyzeVideoTool = defineTool({
   schema: external_exports.object({
     source: external_exports.string().describe("Path to video file"),
     prompt: external_exports.string().describe("Analysis prompt"),
-    model: external_exports.enum(["flash", "flash-2.5", "pro"]).optional().describe("Gemini model variant (default: flash)"),
+    model: external_exports.enum(["flash", "latest", "flash-3", "flash-2.5", "pro", "pro-3.1"]).optional().describe("Gemini model variant (default: flash)"),
     fps: external_exports.number().optional().describe("Frames per second for sampling"),
     start: external_exports.number().optional().describe("Start time in seconds"),
     end: external_exports.number().optional().describe("End time in seconds"),
@@ -475895,7 +476079,7 @@ var analyzeReviewTool = defineTool({
     storyboardPath: external_exports.string().optional().describe("Path to storyboard.json for intent comparison"),
     autoApply: external_exports.boolean().optional().describe("Automatically apply suggested fixes (default: false)"),
     verify: external_exports.boolean().optional().describe("Re-review after applying fixes (default: false)"),
-    model: external_exports.enum(["flash", "flash-2.5", "pro"]).optional().describe("Gemini model variant (default: flash)"),
+    model: external_exports.enum(["flash", "latest", "flash-3", "flash-2.5", "pro", "pro-3.1"]).optional().describe("Gemini model variant (default: flash)"),
     outputPath: external_exports.string().optional().describe("Output path for fixed video")
   }),
   async execute(args) {
@@ -475977,7 +476161,7 @@ var generateMotionTool = defineTool({
       "Analyze the base video with Gemini before generating motion graphics: auto, off, or required (default: auto)"
     ),
     understandingPrompt: external_exports.string().optional().describe("Custom prompt for video understanding when --video is provided"),
-    model: external_exports.enum(["sonnet", "opus", "gemini", "gemini-3.1-pro"]).optional().describe("LLM model for code generation (default: sonnet)"),
+    model: external_exports.enum(["sonnet", "opus", "gemini", "gemini-2.5-pro", "gemini-3.1-pro"]).optional().describe("LLM model for code generation (default: sonnet)"),
     output: external_exports.string().optional().describe("Output path (TSX if code-only, MP4 if rendered)")
   }),
   async execute(args) {
@@ -476242,6 +476426,7 @@ var generateVideoTool = defineTool({
       "Video provider (default: seedance when FAL_API_KEY is configured, otherwise first configured provider)"
     ),
     image: external_exports.string().optional().describe("Reference image path for image-to-video"),
+    endImage: external_exports.string().optional().describe("Ending frame image path for Seedance image-to-video"),
     refImages: external_exports.array(external_exports.string()).optional().describe("Reference images for Seedance reference-to-video"),
     refVideos: external_exports.array(external_exports.string()).optional().describe("Reference videos for Seedance reference-to-video"),
     refAudio: external_exports.array(external_exports.string()).optional().describe("Reference audio files for Seedance reference-to-video"),
@@ -476249,7 +476434,7 @@ var generateVideoTool = defineTool({
     ratio: external_exports.string().optional().describe("Aspect ratio: 16:9, 9:16, 1:1 (default: 16:9)"),
     mode: external_exports.string().optional().describe("Kling mode: std or pro"),
     negative: external_exports.string().optional().describe("Negative prompt (Seedance/Kling/Veo)"),
-    resolution: external_exports.string().optional().describe("Resolution: 720p, 1080p, 4k (Veo only)"),
+    resolution: external_exports.string().optional().describe("Resolution: 480p, 720p, 1080p, or 4k depending on provider"),
     veoModel: external_exports.string().optional().describe("Veo model: 3.0, 3.1, 3.1-fast"),
     runwayModel: external_exports.string().optional().describe("Runway model: gen4.5, gen4_turbo"),
     seedanceModel: external_exports.string().optional().describe("Seedance variant: quality or fast (fal.ai only)"),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vibeframe/mcp-server",
-  "version": "0.105.2",
+  "version": "0.106.0",
   "description": "VibeFrame MCP Server - AI-native video editing via Model Context Protocol",
   "type": "module",
   "bin": {
@@ -57,8 +57,8 @@
     "tsx": "^4.21.0",
     "typescript": "^5.3.3",
     "vitest": "^1.2.2",
-    "@vibeframe/cli": "0.105.2",
-    "@vibeframe/core": "0.105.2"
+    "@vibeframe/cli": "0.106.0",
+    "@vibeframe/core": "0.106.0"
   },
   "engines": {
     "node": ">=20"