@lightcone-ai/daemon 0.15.26 → 0.15.27

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lightcone-ai/daemon",
-  "version": "0.15.26",
+  "version": "0.15.27",
   "type": "module",
   "main": "src/index.js",
   "bin": {

package/src/_vendor/video/composer/index.js

@@ -4,6 +4,10 @@ import { access, mkdir, mkdtemp, rm, stat, writeFile } from 'node:fs/promises';
 import { constants as fsConstants } from 'node:fs';
 import os from 'node:os';
 
+const SUBTITLE_FONT = 'WenQuanYi Micro Hei';
+const SUBTITLE_FONT_SIZE = 72;
+const SUBTITLE_MARGIN_V = 80;
+
 const MAX_STDERR_LENGTH = 4000;
 
 const TRANSCODE_TARGETS = Object.freeze({
@@ -286,10 +290,46 @@ function resolveTranscodeTarget(target) {
   throw new Error(`unsupported transcode target: ${target}`);
 }
 
+function msToAssTimestamp(ms) {
+  const totalCs = Math.round(Math.max(0, ms) / 10);
+  const cs = totalCs % 100;
+  const totalSec = Math.floor(totalCs / 100);
+  const sec = totalSec % 60;
+  const totalMin = Math.floor(totalSec / 60);
+  const min = totalMin % 60;
+  const hr = Math.floor(totalMin / 60);
+  return `${hr}:${String(min).padStart(2, '0')}:${String(sec).padStart(2, '0')}.${String(cs).padStart(2, '0')}`;
+}
+
+export function buildAssContent(subtitles = [], { playResX = 1080, playResY = 1920 } = {}) {
+  const header = [
+    '[Script Info]',
+    'ScriptType: v4.00+',
+    `PlayResX: ${playResX}`,
+    `PlayResY: ${playResY}`,
+    'WrapStyle: 0',
+    '',
+    '[V4+ Styles]',
+    'Format: Name, Fontname, Fontsize, PrimaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding',
+    `Style: Default,${SUBTITLE_FONT},${SUBTITLE_FONT_SIZE},&H00FFFFFF,&H00000000,&H00000000,-1,0,0,0,100,100,0,0,1,4,0,2,30,30,${SUBTITLE_MARGIN_V},1`,
+    '',
+    '[Events]',
+    'Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text',
+  ].join('\n');
+
+  const events = subtitles.map(({ text, start_ms, end_ms }) => {
+    const safeText = String(text ?? '').replace(/\r?\n/g, '\\N').replace(/,/g, '{\\,}');
+    return `Dialogue: 0,${msToAssTimestamp(start_ms)},${msToAssTimestamp(end_ms)},Default,,0,0,0,,${safeText}`;
+  });
+
+  return `${header}\n${events.join('\n')}\n`;
+}
+
 export async function transcodeForPlatform({
   input,
   output,
   target = 'short_video_cn',
+  subtitlesAssPath = null,
 } = {}) {
   const inputPath = normalizePath(input, 'input');
   await ensureReadableFile(inputPath, 'input');
@@ -301,11 +341,16 @@ export async function transcodeForPlatform({
   await ensureParentDir(outputPath);
 
   const preset = resolveTranscodeTarget(target);
-  const vf = [
+  const vfParts = [
     `scale=${preset.width}:${preset.height}:force_original_aspect_ratio=decrease`,
     `pad=${preset.width}:${preset.height}:(ow-iw)/2:(oh-ih)/2:black`,
     'setsar=1',
-  ].join(',');
+  ];
+  if (subtitlesAssPath) {
+    const escapedPath = subtitlesAssPath.replace(/\\/g, '/').replace(/:/g, '\\:').replace(/'/g, "\\'");
+    vfParts.push(`subtitles='${escapedPath}'`);
+  }
+  const vf = vfParts.join(',');
 
   await runProcess('ffmpeg', [
     '-y',

package/src/chat-bridge.js

@@ -7,6 +7,7 @@ import { createHash, randomUUID } from 'crypto';
 import path, { extname } from 'path';
 import { homedir } from 'os';
 import {
+  buildAssContent,
   concatVideos,
   muxAudioToVideo,
   probeDurationMs,
@@ -1498,7 +1499,7 @@ server.tool('record_url_narration',
 
 // ── compose_video ───────────────────────────────────────────────────────────────
 server.tool('compose_video',
-  'Compose a final short video by muxing audio onto a base video, optionally concatenating an outro, and transcoding to platform spec.',
+  'Compose a final short video by muxing audio onto a base video, optionally burning subtitles, concatenating an outro, and transcoding to platform spec.',
   {
     video_path: z.string().describe('Base silent video path. Relative paths resolve from the current workspace.'),
     audio_segments: z.array(z.object({
@@ -1507,14 +1508,19 @@ server.tool('compose_video',
       phase: z.string().optional().describe('Optional phase id. Used with events_log to derive start time.'),
     })).describe('Ordered or unordered narration audio segments.'),
     events_log: z.array(z.any()).optional().describe('Optional recorder event log. Used to resolve segment start time by phase.'),
+    subtitles: z.array(z.object({
+      text: z.string().describe('Subtitle text for this segment (the narration sentence).'),
+      start_ms: z.number().describe('Subtitle start time in milliseconds.'),
+      end_ms: z.number().describe('Subtitle end time in milliseconds.'),
+    })).optional().describe('Subtitle segments to burn into the video. Pass each phase narration text with its start/end time (derived from detail_sections dwell_ms). Omit to produce no subtitles.'),
     outro_path: z.string().optional().describe('Optional outro mp4 path. If omitted, uses ~/.lightcone/assets/outros/default.mp4 when present.'),
     target: z.enum(['short_video_cn', 'douyin', 'xhs']).optional().describe('Transcode target profile. Defaults to short_video_cn.'),
   },
-  async ({ video_path, audio_segments, events_log, outro_path, target }) => {
+  async ({ video_path, audio_segments, events_log, subtitles, outro_path, target }) => {
     if (isBlockedCvmaxEditorVideoTool('compose_video')) {
       return cvmaxEditorVideoToolError('compose_video');
     }
-    const composeInput = { video_path, audio_segments, events_log, outro_path, target };
+    const composeInput = { video_path, audio_segments, events_log, subtitles, outro_path, target };
     try {
       const result = await runMandatoryLocalTool({
         toolName: 'compose_video',
@@ -1541,7 +1547,16 @@ server.tool('compose_video',
           const muxedPath = path.join(VIDEO_COMPOSE_LOCAL_DIR, `compose-${runId}.muxed.mp4`);
           const concatPath = path.join(VIDEO_COMPOSE_LOCAL_DIR, `compose-${runId}.concat.mp4`);
           const finalPath = path.join(VIDEO_COMPOSE_LOCAL_DIR, `compose-${runId}.final.mp4`);
-          const intermediates = [muxedPath, concatPath];
+          const assPath = path.join(VIDEO_COMPOSE_LOCAL_DIR, `compose-${runId}.ass`);
+          const intermediates = [muxedPath, concatPath, assPath];
+
+          const subtitleSegments = Array.isArray(checkedInput.subtitles) ? checkedInput.subtitles : [];
+          let subtitlesAssPath = null;
+          if (subtitleSegments.length > 0) {
+            const assContent = buildAssContent(subtitleSegments);
+            writeFileSync(assPath, assContent, 'utf8');
+            subtitlesAssPath = assPath;
+          }
 
           try {
             await muxAudioToVideo({
@@ -1564,6 +1579,7 @@ server.tool('compose_video',
               input: composedPath,
               output: finalPath,
               target: targetProfile,
+              subtitlesAssPath,
             });
 
             const durationMs = await probeDurationMs(finalPath);
@@ -1573,6 +1589,7 @@ server.tool('compose_video',
               durationMs,
               outroPath: resolvedOutroPath,
               target: targetProfile,
+              subtitles: subtitleSegments.length > 0,
             };
           } catch (error) {
             cleanupLocalFiles([...intermediates, finalPath]);
@@ -1590,6 +1607,7 @@ server.tool('compose_video',
             `final_video_path=${result.finalVideoPath}\n` +
             `duration_ms=${result.durationMs}\n` +
             `target=${result.target}\n` +
+            `subtitles=${result.subtitles ? 'burned' : 'none'}\n` +
             `outro=${outroText}`,
         }],
       };