feat: make intake auto extract copy

api/main.py

@@ -3785,6 +3785,16 @@ class RewriteStoryboardScriptReq(BaseModel):
     segments: list[ScriptRewriteSegmentReq] = Field(default_factory=list)
 
 
+class UpdateAudioScriptReq(BaseModel):
+    source_text: str | None = None
+    source_zh: str | None = None
+    rewritten_text: str | None = None
+    rewritten_text_zh: str | None = None
+    speaker_profile: str | None = None
+    rhythm_profile: str | None = None
+    background_audio_profile: str | None = None
+
+
 _TRANSLATION_CACHE: dict[str, str] = {}
 
 
@@ -4272,6 +4282,33 @@ async def trigger_transcribe(job_id: str, bg: BackgroundTasks) -> Job:
     return job_with_artifacts(job)
 
 
+@app.patch("/jobs/{job_id}/audio-script", response_model=Job)
+def update_audio_script(job_id: str, req: UpdateAudioScriptReq) -> Job:
+    job = JOBS.get(job_id)
+    if not job:
+        raise HTTPException(404, "job not found")
+    audio_script = job.audio_script or AudioScript()
+    patch: dict[str, str] = {}
+    for field in (
+        "source_text",
+        "source_zh",
+        "rewritten_text",
+        "rewritten_text_zh",
+        "speaker_profile",
+        "rhythm_profile",
+        "background_audio_profile",
+    ):
+        value = getattr(req, field)
+        if value is not None:
+            patch[field] = str(value).strip()
+    if not patch:
+        return job_with_artifacts(job)
+    patch["status"] = "completed"
+    patch["created_at"] = audio_script.created_at or time.time()
+    update(job, audio_script=audio_script.model_copy(update=patch))
+    return job_with_artifacts(job)
+
+
 @app.get("/jobs/{job_id}/video.mp4")
 def get_video(job_id: str):
     p = job_dir(job_id) / "source.mp4"

web/app/page.tsx

@@ -229,7 +229,7 @@ export default function Home() {
       const created = await uploadJob(file)
       addJob(created)
       setProductionJobIds((prev) => new Set(prev).add(created.id))
-      toast.success(`已上传 ${created.id.slice(0, 8)}，视频就绪后自动跑音频和抽帧`)
+      toast.success(`已上传 ${created.id.slice(0, 8)}，视频就绪后自动提取音频文案`)
     } catch (e) {
       toast.error("上传失败：" + (e instanceof Error ? e.message : String(e)))
     } finally {
@@ -473,31 +473,13 @@ export default function Home() {
       try {
         const updated = await triggerTranscribe(target.id)
         updateJobInList(updated)
-        toast.info("音频路已启动：字幕、讲话人、节奏和背景音同步解析")
+        toast.info("文案路已启动：字幕、讲话人、节奏和背景音同步解析")
       } catch (e) {
         autoTriggeredRef.current.delete(audioKey)
         toast.error("音频解析启动失败：" + (e instanceof Error ? e.message : String(e)))
       }
     }
-
-    const visualKey = `${target.id}:visual`
-    const hasVisualResult = target.frames.length > 0
-    const visualRunning = target.status === "splitting"
-    if (!hasVisualResult && !visualRunning && !autoTriggeredRef.current.has(visualKey)) {
-      autoTriggeredRef.current.add(visualKey)
-      const frameTarget = frameTargets[target.id] ?? "motion"
-      const frameCount = frameCounts[target.id] ?? 12
-      const frameQuality = frameQualities[target.id] ?? "accurate"
-      try {
-        const updated = await analyzeJob(target.id, frameCount, frameTarget, "replace", frameQuality)
-        updateJobInList(updated)
-        toast.info(`视觉路已启动：${FRAME_QUALITY_LABELS[frameQuality]} · ${FRAME_TARGET_LABELS[frameTarget]} · ${frameCount} 张参考帧`)
-      } catch (e) {
-        autoTriggeredRef.current.delete(visualKey)
-        toast.error("视觉抽帧启动失败：" + (e instanceof Error ? e.message : String(e)))
-      }
-    }
-  }, [frameCounts, frameQualities, frameTargets, updateJobInList])
+  }, [updateJobInList])
 
   const ensureDefaultProductRefs = useCallback(async (jobId: string) => {
     const cached = defaultProductRefsByJob[jobId]
@@ -577,21 +559,20 @@ export default function Home() {
     }
     if (!created && target.status === "failed") {
       autoTriggeredRef.current.delete(`${target.id}:audio`)
-      autoTriggeredRef.current.delete(`${target.id}:visual`)
     }
     if (!created && target.status === "failed" && !target.video_url) {
       try {
         target = await retryJobDownload(target.id)
         updateJobInList(target)
-        toast.info("已重新提交下载；下载完成后会自动跑音频文案路和视觉抽帧路")
+        toast.info("已重新提交下载；下载完成后会自动跑音频文案路")
       } catch (e) {
         toast.error("重新下载失败：" + (e instanceof Error ? e.message : String(e)))
         return
       }
     }
     setProductionJobIds((prev) => new Set(prev).add(target.id))
-    if (target.video_url) toast.success("已进入并行素材分析：音频文案路和视觉抽帧路会同步推进")
-    else toast.success("已进入并行素材分析：下载完成后自动跑音频文案路和视觉抽帧路")
+    if (target.video_url) toast.success("已进入第一步：自动提取音频文案")
+    else toast.success("已进入第一步：下载完成后自动提取音频文案")
     void startProductionLanesForJob(target)
   }, [handleSubmit, job, startProductionLanesForJob, updateJobInList])
 

web/components/ad-recreation-board.tsx

@@ -61,6 +61,7 @@ import {
   sourceAudioUrl,
   subjectTemplateImageUrl,
   updateElement,
+  updateAudioScript,
   updateStoryboard,
   uploadStoryboardAsset,
   translateText,
@@ -1836,13 +1837,6 @@ export function AdRecreationBoard({
     }
   }, [])
 
-  const submitUrl = () => {
-    const trimmed = url.trim()
-    if (!trimmed) return
-    data.onSubmitUrl(trimmed)
-    setUrl("")
-  }
-
   const startProduction = () => {
     const trimmed = url.trim()
     data.onStartProduction?.(trimmed || undefined)
@@ -2079,8 +2073,8 @@ export function AdRecreationBoard({
             url={url}
             setUrl={setUrl}
             fileRef={fileRef}
-            onSubmitUrl={submitUrl}
             onStartProduction={startProduction}
+            runtimeModels={runtimeModels}
           />
 
           <section className="skg-board-panel flex min-h-0 flex-col rounded-lg border border-white/10 bg-white/[0.035] shadow-2xl">
@@ -2090,10 +2084,10 @@ export function AdRecreationBoard({
                   <div className="flex items-center gap-2">
                     <span className="inline-flex h-7 w-7 items-center justify-center rounded-md border border-[#d6b36a]/18 bg-[#d6b36a]/10 text-[#f2d58a]"><Mic className="h-3.5 w-3.5" /></span>
                     <WorkflowStepBadge step={workflow.source} compact />
-                    <h2 className="text-[15px] font-semibold leading-tight text-white">源视频解析与参考帧</h2>
+                    <h2 className="text-[15px] font-semibold leading-tight text-white">源视频文案提取</h2>
                   </div>
                   <div className="mt-1 truncate text-[11px] text-white/38" title={statusMessage}>
-                    {statusMessage || "下载源视频后解析音频，再抽参考帧并生成相似主体。"}
+                    {statusMessage || "导入素材后自动下载并提取音频文案；抽帧参考保留为手动工具。"}
                   </div>
                 </div>
                 <div className="flex shrink-0 items-center gap-2">
@@ -2111,7 +2105,6 @@ export function AdRecreationBoard({
                   <Requirement label="视频" ready={!!job?.video_url} detail={job?.status === "downloading" ? "下载中" : job?.video_url ? "已就绪" : "待下载"} />
                   <Requirement label="音频" ready={!!job?.source_audio_url} detail={audioRunning ? "解析中" : job?.source_audio_url ? "已提取" : "待提取"} />
                   <Requirement label="文案" ready={audioReady} detail={audioReady ? `${transcriptCount} 段` : "待解析"} />
-                  <Requirement label="参考帧" ready={visualReady} detail={visualRunning ? "抽帧中" : visualReady ? `${job?.frames.length ?? 0} 张` : "待抽帧"} />
                 </div>
 
                 <details className="group rounded-md border border-white/10 bg-black/28 p-2">
@@ -2179,8 +2172,8 @@ function MaterialColumn({
   url,
   setUrl,
   fileRef,
-  onSubmitUrl,
   onStartProduction,
+  runtimeModels,
 }: {
   data: NodeData
   step: WorkflowStep
@@ -2190,12 +2183,12 @@ function MaterialColumn({
   url: string
   setUrl: (value: string) => void
   fileRef: RefObject<HTMLInputElement | null>
-  onSubmitUrl: () => void
   onStartProduction: () => void
+  runtimeModels?: RuntimeModels
 }) {
   const actionLabel = !url.trim() && job?.status === "failed"
-    ? job.video_url ? "重新解析" : "重新下载"
-    : "开始分析"
+    ? job.video_url ? "重新提文案" : "重新下载"
+    : "导入并提文案"
   return (
     <section className="skg-board-panel flex min-h-0 flex-col gap-3 rounded-lg border border-white/10 bg-white/[0.035] p-3 shadow-2xl">
       <header className="shrink-0 border-b border-white/10 pb-3">
@@ -2204,14 +2197,14 @@ function MaterialColumn({
           <WorkflowStepBadge step={step} compact />
         </div>
         <h2 className="text-[15px] font-semibold leading-tight text-white">素材输入</h2>
-        <p className="mt-1 text-[12px] leading-snug text-white/42">一个素材就是一次文件任务</p>
+        <p className="mt-1 text-[12px] leading-snug text-white/42">导入后自动下载、提取音频并生成可改文案</p>
       </header>
 
       <div className="flex gap-2">
         <input
           value={url}
           onChange={(e) => setUrl(e.target.value)}
-          onKeyDown={(e) => { if (e.key === "Enter") onSubmitUrl() }}
+          onKeyDown={(e) => { if (e.key === "Enter") onStartProduction() }}
           placeholder="粘贴 TK / 信息流视频链接"
           className="h-10 min-w-0 flex-1 rounded-md border border-white/10 bg-black/45 px-3 text-[13px] text-white outline-none placeholder:text-white/28 focus:border-[#d6b36a]/60"
         />
@@ -2245,6 +2238,13 @@ function MaterialColumn({
         />
       </div>
 
+      <MaterialScriptEditor
+        job={job}
+        onJobUpdate={data.onJobUpdate}
+        onTranscribe={() => data.onTranscribeAudio?.(job?.id)}
+        runtimeModels={runtimeModels}
+      />
+
       <div className="min-h-0 flex-1 space-y-2 overflow-y-auto pr-1">
         {jobs.length ? jobs.map((item, index) => (
           <MaterialCard
@@ -2256,13 +2256,167 @@ function MaterialColumn({
             onDelete={data.onDeleteJob ? () => data.onDeleteJob?.(item.id) : undefined}
           />
         )) : (
-          <EmptyState text="还没有素材。每导入一个链接或上传一个文件，就会新增一个素材任务。" />
+          <EmptyState text="还没有素材。每导入一个链接或上传一个文件，就会新增一个素材任务，并自动提取音频文案。" />
         )}
       </div>
     </section>
   )
 }
 
+function MaterialScriptEditor({
+  job,
+  onJobUpdate,
+  onTranscribe,
+  runtimeModels,
+}: {
+  job: Job | null
+  onJobUpdate: (job: Job) => void
+  onTranscribe?: () => Promise<void> | void
+  runtimeModels?: RuntimeModels
+}) {
+  const script = job?.audio_script
+  const [draftZh, setDraftZh] = useState("")
+  const [draftEn, setDraftEn] = useState("")
+  const [saving, setSaving] = useState(false)
+  const [optimizing, setOptimizing] = useState(false)
+  const lastOptimizedZh = useRef("")
+
+  useEffect(() => {
+    const nextZh = script?.rewritten_text_zh?.trim() || script?.source_zh?.trim() || ""
+    const nextEn = script?.rewritten_text?.trim() || script?.source_text?.trim() || ""
+    setDraftZh(nextZh)
+    setDraftEn(nextEn)
+    lastOptimizedZh.current = nextZh
+  }, [job?.id, script?.rewritten_text, script?.rewritten_text_zh, script?.source_text, script?.source_zh])
+
+  const processing = !!job && (
+    job.status === "created" ||
+    job.status === "downloading" ||
+    job.status === "transcribing" ||
+    script?.status === "rewriting"
+  )
+  const hasScript = !!(script?.source_text?.trim() || script?.source_zh?.trim() || draftEn.trim() || draftZh.trim())
+  const canSave = !!job && !!(draftEn.trim() || draftZh.trim())
+
+  const saveDraft = async (nextEn = draftEn, nextZh = draftZh, quiet = false) => {
+    if (!job) return false
+    setSaving(true)
+    try {
+      const updated = await updateAudioScript(job.id, {
+        rewritten_text: nextEn.trim(),
+        rewritten_text_zh: nextZh.trim(),
+      })
+      onJobUpdate(updated)
+      if (!quiet) toast.success("文案已保存到当前素材")
+      return true
+    } catch (e) {
+      toast.error("保存文案失败：" + (e instanceof Error ? e.message : String(e)))
+      return false
+    } finally {
+      setSaving(false)
+    }
+  }
+
+  const optimizeChinese = async () => {
+    if (!job) return
+    const zh = draftZh.trim()
+    if (!zh || !containsCjk(zh) || zh === lastOptimizedZh.current) return
+    setOptimizing(true)
+    try {
+      const english = await translateText(zh, "en")
+      const nextEn = english.trim() || draftEn
+      setDraftEn(nextEn)
+      lastOptimizedZh.current = zh
+      const saved = await saveDraft(nextEn, zh, true)
+      if (saved) toast.success("中文已同步优化成英文")
+    } catch (e) {
+      toast.error("中文优化英文失败：" + (e instanceof Error ? e.message : String(e)))
+    } finally {
+      setOptimizing(false)
+    }
+  }
+
+  return (
+    <section className="rounded-md border border-white/10 bg-black/30 p-2.5">
+      <div className="mb-2 flex items-center justify-between gap-2">
+        <SectionTitle icon={<FileText className="h-4 w-4" />} title="文案确认稿" />
+        <div className="flex shrink-0 items-center gap-1.5">
+          {runtimeModels ? <ModelTrace trace={audioModelTrace(runtimeModels)} compact /> : null}
+          <button
+            type="button"
+            onClick={() => onTranscribe?.()}
+            disabled={!job?.video_url || processing}
+            className="inline-flex h-7 items-center gap-1 rounded-md border border-white/10 bg-white/[0.04] px-2 text-[10.5px] font-semibold text-white/58 transition hover:border-white/20 hover:bg-white/[0.07] disabled:cursor-not-allowed disabled:opacity-40"
+          >
+            {processing ? <Loader2 className="h-3 w-3 animate-spin" /> : <Mic className="h-3 w-3" />}
+            重提
+          </button>
+        </div>
+      </div>
+
+      {!job ? (
+        <div className="rounded border border-dashed border-white/10 bg-black/22 p-3 text-[11px] leading-relaxed text-white/38">
+          粘贴链接或上传视频后，这里会自动显示原音频文案和可编辑的新广告文案。
+        </div>
+      ) : !hasScript ? (
+        <div className="rounded border border-white/10 bg-black/22 p-3 text-[11px] leading-relaxed text-white/42">
+          {processing ? "正在下载 / 提取音频 / 生成文案，完成后会自动出现在这里。" : job.video_url ? "视频已就绪，等待音频文案。可点“重提”手动启动。" : "等待视频下载完成后自动提取音频文案。"}
+        </div>
+      ) : (
+        <div className="space-y-2">
+          <label className="block">
+            <div className="mb-1 flex items-center justify-between text-[10.5px] text-white/44">
+              <span>中文确认稿</span>
+              <span>{optimizing ? "正在优化英文" : "改完中文会自动同步英文"}</span>
+            </div>
+            <textarea
+              value={draftZh}
+              onChange={(e) => setDraftZh(e.target.value)}
+              onBlur={() => void optimizeChinese()}
+              rows={4}
+              className="min-h-[86px] w-full resize-y rounded-md border border-white/10 bg-black/45 px-2.5 py-2 text-[12px] leading-relaxed text-white outline-none placeholder:text-white/25 focus:border-[#d6b36a]/55"
+              placeholder="这里改中文文案"
+            />
+          </label>
+
+          <label className="block">
+            <div className="mb-1 text-[10.5px] text-white/44">英文生成稿（后续视频 prompt 用这个）</div>
+            <textarea
+              value={draftEn}
+              onChange={(e) => setDraftEn(e.target.value)}
+              rows={3}
+              className="min-h-[72px] w-full resize-y rounded-md border border-white/10 bg-black/45 px-2.5 py-2 font-mono text-[11px] leading-relaxed text-white/76 outline-none placeholder:text-white/25 focus:border-[#d6b36a]/55"
+              placeholder="English copy for generation"
+            />
+          </label>
+
+          <div className="flex items-center justify-between gap-2">
+            <details className="group min-w-0 flex-1 rounded border border-white/10 bg-black/22 px-2 py-1.5">
+              <summary className="flex cursor-pointer list-none items-center justify-between gap-2 text-[10.5px] text-white/44">
+                <span>原音频文案</span>
+                <ChevronDown className="h-3.5 w-3.5 transition group-open:rotate-180" />
+              </summary>
+              <div className="mt-1 max-h-24 overflow-y-auto text-[10.5px] leading-relaxed text-white/48">
+                {script?.source_zh ? <p>{script.source_zh}</p> : null}
+                {script?.source_text ? <p className="mt-1 font-mono text-white/34">{script.source_text}</p> : null}
+              </div>
+            </details>
+            <button
+              type="button"
+              disabled={!canSave || saving || optimizing}
+              onClick={() => void saveDraft()}
+              className="skg-primary-action inline-flex h-8 shrink-0 items-center gap-1.5 px-2.5 text-[11px] font-semibold transition disabled:cursor-not-allowed disabled:opacity-45"
+            >
+              {saving || optimizing ? <Loader2 className="h-3.5 w-3.5 animate-spin" /> : <Check className="h-3.5 w-3.5" />}
+              保存
+            </button>
+          </div>
+        </div>
+      )}
+    </section>
+  )
+}
+
 function AudioIntakePanel({
   job,
   selectedFrames,

web/lib/api.ts

@@ -1044,6 +1044,25 @@ export async function triggerTranscribe(id: string): Promise<Job> {
   return res.json()
 }
 
+export async function updateAudioScript(
+  id: string,
+  body: Partial<Pick<AudioScript,
+    "source_text" | "source_zh" | "rewritten_text" | "rewritten_text_zh" |
+    "speaker_profile" | "rhythm_profile" | "background_audio_profile"
+  >>,
+): Promise<Job> {
+  const res = await fetch(`${API_BASE}/jobs/${id}/audio-script`, {
+    method: "PATCH",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify(body),
+  })
+  if (!res.ok) {
+    const txt = await res.text().catch(() => "")
+    throw new Error(`updateAudioScript ${res.status} ${txt.slice(0, 200)}`)
+  }
+  return res.json()
+}
+
 export async function analyzeJob(
   id: string,
   frames = 12,