@amg555/frameforge-mcp 2.7.8 → 2.7.9

package/.claude-plugin/marketplace.json

@@ -2,7 +2,7 @@
   "name": "frameforge",
   "displayName": "FrameForge — AI Video Production",
   "description": "Full AI video production engine. Produce short-form reels (1080×1920), YouTube videos (1920×1080), or code-driven animated explainers — directly from your IDE. Playwright + FFmpeg renders every frame. 34 MCP tools. Whisper, yt-dlp, broadcast audio, 24 film styles, ClaudeAnimationBase, onetake 39-move library, HyperFrames motion craft. 62 Motion Graphics Styles.",
-  "version": "2.7.8",
+  "version": "2.7.9",
   "homepage": "https://github.com/AMG555/FrameForge",
   "plugins": [
     {

package/CHANGELOG.md

@@ -4,6 +4,17 @@ All notable changes to FrameForge are documented here.
 
 ---
 
+## [2.7.9] — 2026-11-13
+
+### Fixed
+- **TTS language reliability**: `selectNarratorVoice` now accepts a `langHint` param (BCP-47) and covers 25 languages with the correct edge-tts Neural voice. Added Unicode-range detection for Korean (`\uAC00-\uD7AF`), Kannada (`\u0C80-\u0CFF`), Bengali (`\u0980-\u09FF`), and Cyrillic/Russian (`\u0400-\u04FF`). Punjabi entry corrected to gender-neutral `pa-IN-OjasvNeural` (edge-tts only ships one Punjabi voice).
+- **Language propagated end-to-end**: `render_tts` tool now accepts a `language` field; `runAssets` passes `sceneLang` to every `tts.synthesize()` call; `_edgeTTS` defensive `safeVoice` fallback now uses `_overrides.language`. `TTSProvider` constructor JSDoc documents `opts.language`.
+- **Voice Consistency Lock**: `runAssets` pre-resolves a single canonical narrator voice across all narration text before the scene loop — no more mid-video voice drift from punctuation/script variations. Per-scene `scene.voice` still overrides for intentional multi-speaker dialogue.
+- **Parallelism default**: `runAssets` function default raised from 1 → 4 (matching `create_video_from_script` and `render_from_scenes`); TTS now runs 4 scenes concurrently by default.
+- **TTS retry delay**: `baseDelayMs` in `runAssets` reduced from 8000ms → 2000ms; inter-batch pause skipped entirely for `edge-tts` (local subprocess, no rate limit).
+
+---
+
 ## [2.7.8] — 2026-11-12
 
 ### Changed

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@amg555/frameforge-mcp",
-  "version": "2.7.8",
+  "version": "2.7.9",
   "description": "FrameForge — Universal AI video creation & autonomous editing engine for any AI agent, CLI, or IDE. npx @amg555/frameforge-mcp. FFmpeg + TTS on your machine. 34 MCP tools. Whisper transcription, yt-dlp ingestion, broadcast-grade audio, 62 procedural motion styles, 6 viral faceless shorts engines, autonomous video editor with retake/silence cutting, J/L-cuts, timeline export to DaVinci/Premiere.",
   "publishConfig": {
     "access": "public"

package/src/mcp-server/server.js

@@ -166,16 +166,17 @@ const TOOLS = {
         sceneIndex:   { type: 'number' },
         text:         { type: 'string' },
         voice:        { type: 'string', description: 'Voice ID (e.g. en-US-GuyNeural). Overrides TTS_VOICE env var.' },
+        language:     { type: 'string', description: 'BCP-47 language hint (e.g. "ml", "hi", "fr"). Auto-selects best voice when voice is not set.' },
       },
     },
-    handler: async ({ productionId, sceneIndex, text, voice }) => {
+    handler: async ({ productionId, sceneIndex, text, voice, language = '' }) => {
       const vdir      = path.join(FILES_DIR, 'video_tmp');
       const outputPath = path.join(vdir, `voice_${sceneIndex}.m4a`);
       await fs.mkdir(vdir, { recursive: true });
 
       await retryPol.executeStage('TTS', async () => {
-        // voice arg overrides env; rate is read from TTS_RATE env inside synthesize()
-        return tts.synthesize({ text, voice, outputPath });
+        // voice arg overrides env; language enables BCP-47 auto-voice selection
+        return tts.synthesize({ text, voice, outputPath, language });
       });
 
       const stat = await fs.stat(outputPath);
@@ -2124,7 +2125,7 @@ Professional audio rules applied:
       planMd += `- L-cuts: current audio extends 0.3s into next clip's video\n\n`;
       planMd += `## 🚀 Execution & Approval\n\n`;
       planMd += `Review this plan, then execute with \`edit_video\` using the payload in \`edit.json\`.\n`;
-      planMd += `\nGenerated by FrameForge v2.7.8 — plan_video_edit with J/L-cut support (lemo-opuscar DIRECTOR.md)\n`;
+      planMd += `\nGenerated by FrameForge v2.7.9 — plan_video_edit with J/L-cut support (lemo-opuscar DIRECTOR.md)\n`;
 
       // 8. Build TREATMENT.md (lemo-opuscar DIRECTOR.md — 8-section format)
       let treatmentMd = '';
@@ -2165,7 +2166,7 @@ Professional audio rules applied:
         treatmentMd += `## 8. Director Notes\n\n`;
         treatmentMd += `_Add creative direction, mood references, or client notes here._\n\n`;
         treatmentMd += `---\n\n`;
-        treatmentMd += `*Generated by FrameForge v2.7.8 — lemo-opuscar DIRECTOR.md format*\n`;
+        treatmentMd += `*Generated by FrameForge v2.7.9 — lemo-opuscar DIRECTOR.md format*\n`;
       }
 
       // 9. Write outputs
@@ -2246,7 +2247,7 @@ Professional audio rules applied:
     handler: async () => {
       const syncFs = require('fs');
       const { MANIFEST_PATH, REPOSITORIES } = require('../skills/sync-upstream');
-      let manifest = { lastSync: null, repos: {}, version: '2.7.8', pendingUpdates: [] };
+      let manifest = { lastSync: null, repos: {}, version: '2.7.9', pendingUpdates: [] };
       if (syncFs.existsSync(MANIFEST_PATH)) {
         try { manifest = JSON.parse(syncFs.readFileSync(MANIFEST_PATH, 'utf8')); } catch (_) {}
       }
@@ -2293,7 +2294,7 @@ Professional audio rules applied:
       return {
         success: true,
         lastSync: manifest.lastSync,
-        manifestVersion: manifest.version || '2.7.8',
+        manifestVersion: manifest.version || '2.7.9',
         totalRepos: repos.length,
         syncedCount,
         pendingUpdates,
@@ -3536,7 +3537,7 @@ Get a style id from list_motion_styles first. Source: FrameForge Motion Library
 
 const SERVER_INFO = {
   name:    'video-pipeline',
-  version: '2.7.8',
+  version: '2.7.9',
 };
 
 const CAPABILITIES = {

package/src/pipeline/providers.js

@@ -370,7 +370,10 @@ class TTSProvider {
    * picks by gender preference (TTS_VOICE_GENDER=male|female|default).
    *
    * @param {string} [explicitVoice]  voice passed by caller (takes priority)
-   * @param {string} [text]           narration text for language detection
+   * @param {string} [text]           narration text — Unicode range detection selects script-appropriate voice
+   * @param {string} [langHint]       BCP-47 language code (e.g. 'fr', 'pt') — used when text is Latin-script
+   *                                  but the production language is known. Falls back to _overrides.language
+   *                                  and then to TTS_LANGUAGE env var.
    * @returns {string}
    */
   selectNarratorVoice(explicitVoice, text = '', langHint = '') {
@@ -430,7 +433,7 @@ class TTSProvider {
         bn: gender === 'female' ? 'bn-IN-TanishaaNeural' : 'bn-IN-BashkarNeural',
         mr: gender === 'female' ? 'mr-IN-AarohiNeural'   : 'mr-IN-ManoharNeural',
         gu: gender === 'female' ? 'gu-IN-DhwaniNeural'   : 'gu-IN-NiranjanNeural',
-        pa: gender === 'female' ? 'pa-IN-OjasvNeural'    : 'pa-IN-OjasvNeural',
+        pa: 'pa-IN-OjasvNeural', // edge-tts only has one Punjabi voice (gender-neutral)
         ur: gender === 'female' ? 'ur-PK-UzmaNeural'     : 'ur-PK-AsadNeural',
         zh: gender === 'female' ? 'zh-CN-XiaoxiaoNeural' : 'zh-CN-YunjianNeural',
         ja: gender === 'female' ? 'ja-JP-NanamiNeural'   : 'ja-JP-KeitaNeural',
@@ -552,7 +555,8 @@ class TTSProvider {
 
     const rawPath   = outputPath.replace(/\.[^.]+$/, '_raw.mp3');
     const txtPath   = rawPath + '.txt';
-    const safeVoice = voice || this.selectNarratorVoice(null, text);
+    // voice is always pre-resolved by synthesize() — this fallback is purely defensive
+    const safeVoice = voice || this.selectNarratorVoice(null, text, this._overrides.language || '');
     // No artificial rate manipulation — use the voice's natural cadence.
     // Only a mild rate adjustment from env (e.g. TTS_RATE='+0%') is supported.
     const rate = this._overrides.rate || process.env.TTS_RATE || '+0%';

package/src/skills/sync-upstream.js

@@ -117,7 +117,7 @@ function fetchJson(url) {
   return new Promise((resolve, reject) => {
     const opts = {
       headers: {
-        'User-Agent': 'FrameForge-Skill-Sync/2.7.8',
+        'User-Agent': 'FrameForge-Skill-Sync/2.7.9',
         'Accept': 'application/vnd.github.v3+json',
       },
     };
@@ -188,7 +188,7 @@ async function syncSkills(opts = {}) {
   const updatedResults = results.filter(r => r.updated);
 
   manifest.lastSync = new Date().toISOString();
-  manifest.version = '2.7.8';
+  manifest.version = '2.7.9';
   manifest.updatedCount = updatedCount;
   if (updatedResults.length > 0) {
     manifest.pendingUpdates = updatedResults.map(r => {
@@ -278,7 +278,7 @@ function getUpstreamSummary() {
   return {
     repoCount,
     lastSync: manifest.lastSync,
-    version: manifest.version || '2.7.8',
+    version: manifest.version || '2.7.9',
     pendingUpdates: pending,
     hasUpdates: pending.length > 0,
     repos: manifest.repos || {},

package/src/workflows/assets.js

@@ -27,7 +27,7 @@ const VDIR      = `${FILES_DIR}/video_tmp`;
 async function runAssets({
   productionId,
   scenes,
-  parallelism = 1,
+  parallelism = 4,
   skipTts = false,
   voice = null,
   topic = '',