tuna-agent 0.1.156 → 0.1.158

package/dist/daemon/analyze-video-handler.js

@@ -144,6 +144,22 @@ function detectPlatform(url) {
         return 'facebook';
     return 'other';
 }
+// yt-dlp `%(title)s` is a clean title on YouTube but the full post caption
+// on TikTok/Douyin/Facebook (often 200+ chars, prefixed with engagement
+// stats). Strip the social-stats prefix, keep the first segment, and cap
+// to a sane title length so the cloned idea doesn't get a paragraph title.
+function cleanSourceTitle(raw) {
+    let t = (raw || '').trim().split('\n')[0].trim();
+    // Drop leading "1K views · 51 reactions | " / "230 likes, 12 comments - " noise.
+    t = t.replace(/^\s*(?:[\d.,]+\s*[KMB]?\s*(?:views?|likes?|reactions?|comments?|shares?|followers?)\b\s*[·,|–\-:]*\s*)+/i, '').trim();
+    // FB/TikTok pack "title | extra | hashtags" — keep the first real segment.
+    const seg = t.split(/\s*[|·]\s*/)[0].trim();
+    if (seg.length >= 8)
+        t = seg;
+    if (t.length > 90)
+        t = t.slice(0, 90).replace(/\s+\S*$/, '').trim() + '…';
+    return t.slice(0, 120);
+}
 // Download a source video across YouTube / TikTok / Douyin / Facebook.
 // yt-dlp supports all of them, but a single rigid `-f` that works for
 // YouTube fails on the others, so try a tolerant 720p-capped format then
@@ -153,6 +169,10 @@ async function downloadSourceVideo(url, dest) {
     const platform = detectPlatform(url);
     const common = [
         '--no-playlist', '--no-warnings', '--retries', '3', '--fragment-retries', '3',
+        // dest ends in .mp4 — force merged/odd containers to mp4 so the file
+        // lands EXACTLY at `dest` (else fs.rename → ENOENT, e.g. YouTube shorts
+        // where bv*+ba merges to .mkv/.webm).
+        '--merge-output-format', 'mp4', '--remux-video', 'mp4',
         '--user-agent', SRC_UA, ...cookieArgs(), '-o', dest, url,
     ];
     const attempts = [
@@ -387,6 +407,15 @@ Rules:
 }
 // Single source of truth for the master-cast prompt block (used by Phase-1
 // and the post-Phase-2 cast reconciliation so the format never drifts).
+//
+// The [CHARACTER CAST LIST] header carries CRITICAL RULES that tell the
+// downstream image generator (Veo3 / FlowKit) how to render the cast. Rule
+// #5 specifically prevents identity-morphing — even when the source video
+// has siblings or look-alike characters, the generated reference sheet
+// must give each character distinctly different facial structures so the
+// per-scene Veo3 renders don't drift into the same face. This rule got
+// lost in a prior refactor; restored to match the reference clone-tool
+// (~/Downloads/if is_character_driven.txt).
 function buildMasterCastPrompt(videoStyle, characters) {
     if (!characters.length)
         return '';
@@ -394,8 +423,15 @@ function buildMasterCastPrompt(videoStyle, characters) {
     const castList = characters.map(c => `- ${c.name}: ${c.description}`).join('\n');
     return (`[AESTHETIC & STYLE]\n${styleLine}\n` +
         `[COMPOSITION & LAYOUT]\nCharacter Reference Sheet. Full-body side-by-side.\n` +
-        `[CHARACTER CAST LIST]\n${castList}\n` +
-        `[TECHNICAL SPECIFICATIONS]\nHigh detail, 8k resolution, consistent facial structures across all frames. Each character has a completely distinct face, hairstyle, body type and age — no two characters look alike.`);
+        `[CHARACTER CAST LIST] (CRITICAL RULES: ` +
+        `1. Max 5 INDIVIDUAL characters. ` +
+        `2. STRICTLY PROHIBIT collective nouns, groups, crowds, or plurals. ` +
+        `3. If a group exists, pick ONLY the most prominent 1 or 2 individuals. ` +
+        `4. Detail unique physical appearance and clothing. ` +
+        `5. FORCE UNIQUE facial and physical structures for EVERY character ` +
+        `(assign distinct body types, face shapes, wrinkles, or accessories) ` +
+        `to prevent identity morphing. List with '- '. English.)\n${castList}\n` +
+        `[TECHNICAL SPECIFICATIONS]\nHigh detail, 8k resolution.`);
 }
 // Post-Phase-2 cast RECONCILE. Phase-1 only sees a 30-frame sample so it can
 // miss a recurring character; the per-scene visionDescribe pass, however,
@@ -522,7 +558,7 @@ export async function analyzeVideo(url, onProgress) {
         let source_title = '';
         try {
             const t = await run(YT_DLP, ['--skip-download', '--no-warnings', '--no-playlist', '--user-agent', SRC_UA, ...cookieArgs(), '--print', '%(title)s', url]);
-            source_title = (t.out || '').trim().split('\n')[0].slice(0, 200);
+            source_title = cleanSourceTitle(t.out);
         }
         catch { /* title is best-effort — analysis still proceeds without it */ }
         progress('Đang tách audio...');

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tuna-agent",
-  "version": "0.1.156",
+  "version": "0.1.158",
   "description": "Tuna Agent - Run AI coding tasks on your machine",
   "bin": {
     "tuna-agent": "dist/cli/index.js"