auto-save 2026-05-14 10:59 (~6)

.memory/worklog.json

@@ -1,26 +1,5 @@
 {
   "entries": [
-    {
-      "files_changed": 1,
-      "hash": "f4a421b",
-      "message": "auto-save 2026-05-13 02:07 (~1)",
-      "ts": "2026-05-13T02:08:10+08:00",
-      "type": "commit"
-    },
-    {
-      "files_changed": 1,
-      "hash": "a63d7c7",
-      "message": "auto-save 2026-05-13 02:13 (~1)",
-      "ts": "2026-05-13T02:14:06+08:00",
-      "type": "commit"
-    },
-    {
-      "files_changed": 1,
-      "hash": "1a5f5be",
-      "message": "auto-save 2026-05-13 02:19 (~1)",
-      "ts": "2026-05-13T02:20:00+08:00",
-      "type": "commit"
-    },
     {
       "files_changed": 1,
       "hash": "bcc7ce0",
@@ -3309,6 +3288,25 @@
       "type": "session-heartbeat",
       "message": "Codex 会话活跃 · 最近命令：codex · 5 项未提交变更 · 最近提交：auto-save 2026-05-14 10:45 (+1, ~5)",
       "files_changed": 5
+    },
+    {
+      "ts": "2026-05-14T10:53:54+08:00",
+      "type": "commit",
+      "message": "auto-save 2026-05-14 10:51 (~7)",
+      "hash": "8bd52f6",
+      "files_changed": 7
+    },
+    {
+      "ts": "2026-05-14T02:56:10Z",
+      "type": "session-heartbeat",
+      "message": "Codex 会话活跃 · 最近命令：codex · 3 项未提交变更 · 最近提交：auto-save 2026-05-14 10:51 (~7)",
+      "files_changed": 3
+    },
+    {
+      "ts": "2026-05-14T02:58:38Z",
+      "type": "session-heartbeat",
+      "message": "Codex 会话活跃 · 最近命令：codex · 5 项未提交变更 · 最近提交：auto-save 2026-05-14 10:51 (~7)",
+      "files_changed": 5
     }
   ]
 }

api/README.md

@@ -18,7 +18,7 @@ uvicorn main:app --port 4291 --reload
 - `GET  /health` — 健康检查 + 配置状态
 - `POST /jobs` `{url}` — 创建 job，后台跑下载/拆轨/抽帧
 - `GET  /jobs/{id}` — 当前状态 + 产物
-- `POST /jobs/{id}/transcribe` — 触发 ASR + 翻译 + SKG 文案改写；配置 MiniMax 后生成配音
+- `POST /jobs/{id}/transcribe` — 触发音频提取 + ASR + 翻译 + SKG 文案改写；配置 MiniMax 后生成配音。前端 Audio 节点提供“提取音频 / 重新提取音频”按钮，不依赖抽帧完成
 - `GET  /jobs/{id}/video.mp4` — 原视频
 - `GET  /jobs/{id}/audio.wav` — 拆轨后的原始音频，供前端底部音频条生成波形
 - `GET  /jobs/{id}/audio-script.mp3` — 改写文案的 MiniMax 配音

api/main.py

@@ -97,7 +97,8 @@ def llm() -> OpenAI:
     return _llm_client
 
 # Pipeline 状态：
-#   created → downloading → downloaded（停，等用户点解析）→ splitting → frames_extracted
+#   created → downloading → downloaded（停，等用户点解析/提取音频）
+#         → splitting → frames_extracted
 #         → transcribing → transcribed | failed
 JobStatus = Literal[
     "created", "downloading", "downloaded",
@@ -1563,7 +1564,17 @@ async def pipeline_transcribe(job_id: str) -> None:
     wav = d / "audio.wav"
     try:
         if not wav.exists():
-            raise RuntimeError("audio.wav 不存在")
+            mp4 = d / "source.mp4"
+            if not mp4.exists():
+                raise RuntimeError("source.mp4 不存在，视频导入完成后再提取音频")
+            update(job, status="transcribing", message="ffmpeg 提取音频轨…", progress=max(job.progress, 45), error="")
+            run([
+                "ffmpeg", "-y", "-i", str(mp4),
+                "-vn", "-ac", "1", "-ar", "16000", "-c:a", "pcm_s16le",
+                str(wav),
+            ])
+            if not wav.exists():
+                raise RuntimeError("音频提取完成但找不到 audio.wav")
 
         if not LLM_API_KEY:
             # 无 key 模式：mock 数据
@@ -2004,9 +2015,12 @@ async def trigger_transcribe(job_id: str, bg: BackgroundTasks) -> Job:
     job = JOBS.get(job_id)
     if not job:
         raise HTTPException(404, "job not found")
-    if job.status not in {"frames_extracted", "transcribed", "failed"}:
+    mp4 = job_dir(job_id) / "source.mp4"
-        raise HTTPException(409, f"status must be frames_extracted/transcribed/failed, got {job.status}")
+    if job.status in {"created", "downloading"} or not mp4.exists():
-    update(job, status="transcribing", progress=max(job.progress, 72), error="", message="准备音频转写…")
+        raise HTTPException(409, f"video not ready, got {job.status}")
+    if job.status in {"splitting", "transcribing"} or job.audio_script.status == "rewriting":
+        raise HTTPException(409, f"job is busy, got {job.status}")
+    update(job, status="transcribing", progress=max(job.progress, 45), error="", message="准备提取音频…")
     bg.add_task(pipeline_transcribe, job_id)
     return job
 

docs/source-analysis.html

@@ -792,7 +792,7 @@ SubjectAsset {
             <tr><td>上传视频</td><td><code>POST /jobs/upload</code></td><td><code>uploadJob</code></td><td>保存 source.mp4，然后同样进入下载完成状态。</td></tr>
             <tr><td>删除输入视频</td><td><code>DELETE /jobs/{id}</code></td><td><code>deleteJob</code></td><td>从任务队列、URL 和磁盘 <code>jobs/&lt;id&gt;</code> 目录移除整个 job，包括源视频、关键帧、元素提取图和生成视频。</td></tr>
             <tr><td>解析视频</td><td><code>POST /jobs/{id}/analyze?frames=&amp;target=&amp;mode=&amp;quality=</code></td><td><code>analyzeJob</code></td><td>拆轨 + 目标化抽关键帧。默认 <code>frames=12</code>；<code>target</code> 支持透明骨架人、综合、清晰主体、转场变化、表情瞬间、动作峰值；当前 UI 默认 <code>transparent_human</code>。透明骨架人目标会先扩大本地候选池，再调用 Vision 按 6 个分数验收；不合格候选自动丢弃并抽下一候选。<code>mode=append</code> 追加新关键帧；<code>quality=auto</code> 根据本机算力和视频时长自动选择快速、精细或极准。多个抽帧请求进入后端队列顺序处理。</td></tr>
-            <tr><td>音频文案轨</td><td><code>POST /jobs/{id}/transcribe</code></td><td><code>triggerTranscribe</code></td><td>读取拆轨得到的 <code>audio.wav</code>，先 ASR 得到英文时间戳段落，再翻译中文，随后按 <code>AUDIO_PRODUCT_BRIEF</code> 生成 <code>audio_script.rewritten_text</code>；配置 <code>MINIMAX_API_KEY</code> 后调用 MiniMax T2A 生成 <code>audio_script.voice_url</code>。</td></tr>
+            <tr><td>音频文案轨</td><td><code>POST /jobs/{id}/transcribe</code></td><td><code>triggerTranscribe</code></td><td>读取拆轨得到的 <code>audio.wav</code>，先 ASR 得到英文时间戳段落，再翻译中文，随后按 <code>AUDIO_PRODUCT_BRIEF</code> 生成 <code>audio_script.rewritten_text</code>；配置 <code>MINIMAX_API_KEY</code> 后调用 MiniMax T2A 生成 <code>audio_script.voice_url</code>。前端在抽帧完成且尚无 transcript 时会自动触发一次；Audio 节点也提供“开始/重新处理音频”按钮。</td></tr>
             <tr><td>原始音频文件</td><td><code>GET /jobs/{id}/audio.wav</code></td><td><code>sourceAudioUrl</code></td><td>返回拆轨得到的 wav；底部 <code>AudioStrip</code> 拉取该文件，用 Web Audio API 解码并计算波形峰值，只读展示，不参与改写。</td></tr>
             <tr><td>改写配音文件</td><td><code>GET /jobs/{id}/audio-script.mp3</code></td><td><code>apiAssetUrl(job.audio_script.voice_url)</code></td><td>返回 MiniMax T2A 生成的 mp3。没有配置 MiniMax 或生成失败时该文件不存在，但改写文案仍会保存在 <code>audio_script.rewritten_text</code>。</td></tr>
             <tr><td>手动加帧</td><td><code>POST /jobs/{id}/frames?t=</code></td><td><code>addManualFrame</code></td><td>按视频时间戳抽一帧，index 递增但 frames 按 timestamp 排序。</td></tr>
@@ -841,7 +841,7 @@ SubjectAsset {
             </tr>
             <tr>
               <td><span class="tag gray">Audio / ASR / Rewrite</span></td>
-              <td>独立声音文案轨：从 <code>audio.wav</code> 提取原始口播、翻译中文、改写成 SKG 产品语境口播；MiniMax T2A 配置后生成配音 mp3。主画布的 <code>AudioNode</code> 用“改前 · 原音频 / 改后 · SKG 口播”摘要展示；底部 <code>AudioStrip</code> 吸附屏幕底端，可拖拽调整高度，按时间段展示英文、中文翻译和波形；侧栏 Rewrite 展开后显示完整审核视图。</td>
+              <td>独立声音文案轨：从 <code>audio.wav</code> 提取原始口播、翻译中文、改写成 SKG 产品语境口播；MiniMax T2A 配置后生成配音 mp3。抽帧完成后自动触发一次，也可在主画布 <code>AudioNode</code> 手动开始/重新处理。<code>AudioNode</code> 用“改前 · 原音频 / 改后 · SKG 口播”摘要展示；底部 <code>AudioStrip</code> 吸附屏幕底端，可拖拽调整高度，按时间段展示英文、中文翻译和波形；侧栏 Rewrite 展开后显示完整审核视图。</td>
               <td>不要阻断视觉素材管线。</td>
               <td><code>AudioNode</code>、<code>AudioStrip</code>、<code>ASRNode</code>、<code>TranslateNode</code>、<code>RewriteNode</code>、<code>pipeline_transcribe</code>、<code>AudioScript</code></td>
             </tr>
@@ -918,6 +918,18 @@ SubjectAsset {
         <h2>变更记录</h2>
         <p>这个记录不是 git log 的替代品。它记录“产品理解发生了什么变化、影响了哪些源码、你以后描述需求时该怎么说”。后续每次改功能都要补一条。</p>
         <div class="changelog">
+          <article class="change">
+            <header>
+              <h3>2026-05-14 · 音频处理支持自动触发和手动重试</h3>
+              <span class="tag gray">Audio</span>
+              <span class="tag blue">Workflow</span>
+            </header>
+            <div class="body">
+              <p><strong>问题：</strong>后端已有 <code>/transcribe</code> 接口，但前端没有入口调用，用户不知道什么时候音频开始工作。</p>
+              <p><strong>改动：</strong>前端在 job 进入 <code>frames_extracted</code> 且没有 transcript 时自动调用一次 <code>triggerTranscribe</code>；<code>AudioNode</code> 增加“开始音频处理 / 重新处理音频”按钮。后端触发接口会立即把 job 状态置为 <code>transcribing</code>，让轮询、节点状态和底部音频条能立刻进入运行态。</p>
+              <p><strong>影响：</strong><code>web/app/page.tsx</code>、<code>web/components/nodes/index.tsx</code>、<code>api/main.py</code>、<code>docs/source-analysis.html</code>。</p>
+            </div>
+          </article>
           <article class="change">
             <header>
               <h3>2026-05-14 · 新增底部可伸缩音频条</h3>

web/app/page.tsx

@@ -390,31 +390,27 @@ export default function Home() {
     if (!targetId) return
     const target = jobs.find((item) => item.id === targetId)
     if (!target) return
-    if (!["frames_extracted", "transcribed", "failed"].includes(target.status)) {
+    if (!target.video_url) {
-      if (!options?.silent) toast.info("先完成抽帧，音频轨会自动开始处理")
+      if (!options?.silent) toast.info("视频导入完成后，可在音频卡片点击提取音频")
+      return
+    }
+    if (target.status === "splitting") {
+      if (!options?.silent) toast.info("当前正在抽帧，结束后可重新点击提取音频")
+      return
+    }
+    if (target.status === "transcribing" || target.audio_script?.status === "rewriting") {
+      if (!options?.silent) toast.info("音频正在处理中")
       return
     }
     try {
       const updated = await triggerTranscribe(targetId)
       updateJobInList(updated)
-      if (!options?.silent) toast.success("音频处理已开始")
+      if (!options?.silent) toast.success("已开始提取音频")
     } catch (e) {
       if (!options?.silent) toast.error("音频处理启动失败：" + (e instanceof Error ? e.message : String(e)))
     }
   }, [activeJobId, jobs, updateJobInList])
 
-  const autoAudioStartedRef = useRef<Set<string>>(new Set())
-  useEffect(() => {
-    for (const item of jobs) {
-      const audioStatus = item.audio_script?.status ?? "idle"
-      const hasAudioOutput = item.transcript.length > 0 || !!item.audio_script?.rewritten_text
-      const ready = item.status === "frames_extracted" && !hasAudioOutput && audioStatus !== "rewriting"
-      if (!ready || autoAudioStartedRef.current.has(item.id)) continue
-      autoAudioStartedRef.current.add(item.id)
-      void handleTranscribeAudio(item.id, { silent: true })
-    }
-  }, [jobs, handleTranscribeAudio])
-
   const handleQuickGenerateVideo = useCallback(async (frameIdx: number, scene: StoryboardScene, model: string) => {
     if (!job) return
     const frame = job.frames.find((f) => f.index === frameIdx)

web/components/nodes/index.tsx

@@ -2109,7 +2109,19 @@ export function AudioNode({ data, selected }: any) {
   const voiceUrl = apiAssetUrl(audioScript?.voice_url)
   const hasASR = transcript.length > 0
   const isRewriting = audioScript?.status === "rewriting"
-  const canTriggerAudio = !!job && ["frames_extracted", "transcribed", "failed"].includes(job.status) && !isRewriting && job.status !== "transcribing"
+  const hasVideo = !!job?.video_url
+  const isAudioBusy = !!job && (job.status === "transcribing" || isRewriting)
+  const isVisualBusy = !!job && job.status === "splitting"
+  const audioButtonDisabled = !job || !hasVideo || isAudioBusy || isVisualBusy
+  const audioButtonLabel = !hasVideo
+    ? "等待视频就绪"
+    : isAudioBusy
+    ? "正在提取音频"
+    : isVisualBusy
+    ? "抽帧中，稍后提取"
+    : hasASR || rewrittenText
+    ? "重新提取音频"
+    : "提取音频"
   const originalPreview = transcript
     .slice(0, 2)
     .map((s) => (s.zh || s.en).trim())
@@ -2139,17 +2151,25 @@ export function AudioNode({ data, selected }: any) {
             {audioScript?.rewrite_model || "AUDIO_REWRITE_MODEL"} → {audioScript?.voice_model || "MiniMax T2A"}
           </span>
         </div>
-        {canTriggerAudio && (
+        {job && (
           <button
             type="button"
+            disabled={audioButtonDisabled}
             onClick={(e) => {
               e.stopPropagation()
+              if (audioButtonDisabled) return
               void d.onTranscribeAudio?.(job.id)
             }}
-            className="inline-flex min-h-8 w-full items-center justify-center gap-1.5 rounded-md border border-violet-300/25 bg-violet-400/10 px-2.5 py-1.5 text-[11px] font-medium text-[var(--text-strong)] transition hover:border-violet-200/45 hover:bg-violet-400/18"
+            className="inline-flex min-h-8 w-full items-center justify-center gap-1.5 rounded-md border border-violet-300/25 bg-violet-400/10 px-2.5 py-1.5 text-[11px] font-medium text-[var(--text-strong)] transition hover:border-violet-200/45 hover:bg-violet-400/18 disabled:cursor-not-allowed disabled:border-white/10 disabled:bg-white/[0.03] disabled:text-[var(--text-faint)]"
           >
-            {hasASR || rewrittenText ? <RotateCcw className="h-3.5 w-3.5" /> : <PlayCircle className="h-3.5 w-3.5" />}
+            {isAudioBusy ? (
-            {hasASR || rewrittenText ? "重新处理音频" : "开始音频处理"}
+              <Loader2 className="h-3.5 w-3.5 animate-spin" />
+            ) : hasASR || rewrittenText ? (
+              <RotateCcw className="h-3.5 w-3.5" />
+            ) : (
+              <PlayCircle className="h-3.5 w-3.5" />
+            )}
+            {audioButtonLabel}
           </button>
         )}
         {(originalPreview || rewrittenText) && (