@starreel/mcp 0.1.83 → 0.1.85

package/SKILL.md

@@ -206,6 +206,15 @@ content that will be rejected.
    character with `set_character_voice` only when you actually need it to
    dub a shot. Always pass that same `voice_id` — never a raw number from
    somewhere else.
+   **In native-audio dramas (`use_clip_audio=true`, the default) binding the
+   customer's own authorized voice is how you lock a character's voice**:
+   videos generated *after* the binding hand that voice to the video vendor as
+   a reference audio, so the character sounds the same in every shot (public
+   library voices are TTS-only and are not sent). Trade-off: that character's
+   shots stop using the portrait-video anchor, so identity rests on the
+   portrait/sheet — have those ready first. Bind **before** `generate_videos`;
+   for videos that already exist, use `repair_episode_dialogue`
+   (`only_flagged=false`) rather than regenerating whole episodes.
 
 6. **Follow the pipeline order — do not skip.** `set_script` → `rewrite_script`
    → `extract_assets` → portraits **+ sheets** → storyboards → frames → videos.
@@ -685,6 +694,40 @@ constraint is already in the prompt and was measured ineffective). The only
 real fix is `regenerate_shot_video` on those shots after switching the drama to
 `seedance-2.5` or `hailuo-3`.
 
+### "动作太慢" / "没演出来" / "像快进" — motion speed inside a shot
+
+Everything here was measured by comparing versions of the same shot. Work
+through it in this order, per shot (`get_shot_prompts` first — you need the
+current `video_prompt` and what the first frame shows):
+
+1. **Check the shot size before the wording.** Wide / full shots cannot show
+   small motion (a face, a hand, an ear): measured motion on wide shots was a
+   fraction of the same batch's medium shots. If the action is small, change
+   `shot_type` to medium or closer (`update_shot`), then `generate_shot_frame`
+   and regenerate the video. Rewording a wide shot was tried twice in the
+   measured case and failed both times.
+2. **Density comes from the number of beats, not verb strength.** Deleting
+   "slow / gently / unhurried" was measured to have **no** effect. What worked:
+   write the `video_prompt` as beats — about **3 beats for 5 seconds, 2 beats
+   for 3 seconds** ("beat 1 … beat 2 … beat 3 …"), so more actually *happens*
+   in those seconds. Adding "one continuous take, no cuts" helps keep the
+   vendor from cutting inside the clip.
+3. **Start from the first frame.** The action must continue from the exact
+   moment the first frame shows. If the prompt's starting state contradicts the
+   frame, the vendor cannot bridge it in one camera move and cuts instead.
+4. **Move beats and `duration` together.** 3 beats squeezed into 3 seconds
+   reads as fast-forward; lengthening a shot without adding beats only makes it
+   slower. Then `regenerate_shot_video` (quote first).
+
+`speed_factor` (`update_shot`, 0.5–2.0, `null` clears) is **playback speed in
+the final cut only** — free, no regeneration, but it does not change what the
+vendor animated, so it cannot fix "nothing happens". Use it for deliberate slow
+motion; avoid it on dialogue shots (the voice speeds up / slows down too).
+
+The platform's default shot breakdown still asks for one continuous action per
+shot. Apply the beat approach shot by shot on complaints; don't rewrite a whole
+episode this way unprompted.
+
 ### "画面里多出一个人 / 多出一件道具" — something appears that shouldn't be there
 
 Two completely different causes with opposite fixes. **Look at the shot's own

package/dist/tools/guide-data.js

@@ -95,8 +95,11 @@ export const ENTRY_POINTS = [
             'set_character_voice(绑到角色;传 voice_id 如 lib:12)',
             'assign_voices',
         ],
-        note: '所有项目默认视频原声(use_clip_audio=true,跳过 TTS);要配音把它设 false,并主动告诉客户可切换。',
-        flow: 'clone_voice → speak_with_voice(试听) → set_character_voice / assign_voices',
+        note: '所有项目默认视频原声(use_clip_audio=true,跳过 TTS);要配音把它设 false,并主动告诉客户可切换。' +
+            '★原声剧里 set_character_voice 绑客户自己的授权音色 = 锁声线:之后出的视频把它交给厂商当参考音频,前后镜声线一致;' +
+            '代价是该角色的镜不再喂定妆视频锚、身份靠定妆图/设定图。只对绑定之后出的视频生效——先绑再出视频;' +
+            '已出好的视频用 repair_episode_dialogue(only_flagged=false)换轨统一,别整集重生。',
+        flow: 'clone_voice → speak_with_voice(试听) → set_character_voice(原声剧:在 generate_videos 之前) / assign_voices',
     },
     {
         customer_has: '歌曲 + 歌词(MV)',
@@ -244,6 +247,16 @@ export const REVIEW_GATE_RULE = '三道闸全部免费、服务端强制(跳过
 export const QA_TOOLS = [
     { symptom: '话没说完就切 / 台词跑到别的镜上', run: 'scan_dialogue_coverage', then: ['repair_episode_dialogue(首选:换音频不重生,便宜)', 'regenerate_shot_video(特写镜或画面也错时)', 'compose_episode'] },
     { symptom: '切太快 / 一个镜头里画面跳来跳去', run: 'scan_intra_shot_cuts', then: ['update_project_settings(video_engine 改 seedance-2.5 或 hailuo-3)', 'regenerate_shot_video'] },
+    // 「动作太慢」的判据全部来自实测(同剧同镜多版对照):删 slow/缓慢这类措辞没有可测效果;
+    // 远景镜的运动量整体只有中景的几分之一;把更多节拍放进同样秒数才真的更有动感。
+    // ★只作为逐镜修法给 agent,平台默认拆镜规则不在这里改(等运动量基线攒够再裁决)。
+    { symptom: '动作太慢 / 没演出来 / 像快进', run: 'get_shot_prompts(先读 video_prompt 与 first_frame_prompt,看首帧画的是哪一刻)', then: [
+            'update_shot(shot_type:远景/全景演不出表情、手部这类细微动作——要看清就改成中景或更近,再 generate_shot_frame 重出首帧;这是分镜问题,改措辞没用)',
+            'update_shot(video_prompt:按节拍写——约 5 秒写 3 拍、3 秒写 2 拍,「第一拍…第二拍…第三拍…」,让这几秒里发生更多事;只删 slow/缓慢/轻轻这类词实测没有效果;动作必须从首帧画的那一刻往后接,起点与首帧矛盾时厂商会自己剪一刀;可加一句「一镜到底、不切镜」)',
+            'update_shot(duration:动作量与时长同向调——3 拍塞进 3 秒就是「像快进」;反过来,只加长不加拍只会更慢)',
+            'regenerate_shot_video(改完 prompt/时长要重出视频才生效,走报价)',
+            'update_shot(speed_factor:只是成片里变速播放,不改厂商生成的内容——做慢镜氛围用,治不了「没演出来」)',
+        ] },
     { symptom: '动作发生在裁剪窗口之外', run: 'recommend_trim_window', then: ['trim_shot'] },
     { symptom: '画面多出一个人 / 多出一件道具', run: 'get_storyboards(先看该镜实际用的首帧)', then: ['generate_shot_frame(首帧本身就有→重生首帧再重生视频)', 'split_shot(帧干净、片中长出来→拆成 3~5 秒短镜)'] },
     { symptom: '出图 / 出视频前想知道哪些镜会被厂商拒', run: 'run_precheck', then: ['update_shot', 'generate_shot_frame'] },

package/dist/tools/produce.js

@@ -1156,7 +1156,7 @@ export function registerProduceTools(server, client) {
         const suffix = qs.toString() ? `?${qs.toString()}` : '';
         return jsonResult(await client.produceGet(`/episodes/${episode_id}/dialogue-coverage${suffix}`));
     });
-    server.tool('update_shot', '逐镜编辑:改单个分镜的文本内容(景别/动作/台词/画面描述/运镜等)与角色绑定(character_ids)。只传要改的字段、其余不动。' +
+    server.tool('update_shot', '逐镜编辑:改单个分镜的文本内容(景别/动作/台词/画面描述/运镜等)、时长(duration)、成片变速(speed_factor)与角色绑定(character_ids)。只传要改的字段、其余不动。' +
         '**免费**(纯文本写库)。★改 dialogue 会自动失效本镜已生成的 TTS 配音与字幕(需重出 tts);' +
         '改文本不会自动重出图/视频,如需让画面跟上文本改动,改完再 regen 对应镜。用 get_storyboards 查改后结果。' +
         '★★原声镜(厂商原生音频)改 dialogue 后,本镜视频会被标记「待重生」——因为台词是**烤进视频人声**的,' +
@@ -1185,6 +1185,8 @@ export function registerProduceTools(server, client) {
         atmosphere: z.string().optional().describe('氛围'),
         director_note: z.string().optional().describe('导演注释'),
         shot_intent: z.string().optional().describe('这镜为什么存在(叙事意图)'),
+        duration: z.number().positive().optional().describe('本镜时长(秒,可带小数如 2.5)。上限=本剧视频引擎的单镜上限,超出直接拒并告知上限(更长内容用 split_shot 拆镜)。有台词时平台按语速律只抬不降,抬了会在回执 speech_duration_note 里说。★改时长不会自动重出视频:已有视频仍是旧长度,要厂商按新时长出就 regenerate_shot_video(报价按新时长算)。★别拿加长治「动作太慢」——同样的动作摊到更长时间里只会更慢,见 qa_tools「动作太慢」'),
+        speed_factor: z.number().positive().nullable().optional().describe('成片变速:<1 慢镜、>1 快放、null 清除(0.5–2.0,越界被钳到范围内)。只在终拼/剪映导出时变速播放(画面 setpts + 声音 atempo,音高不变),**不重出视频、不扣费**,但会改变本镜在成片里占的时长。适合做慢镜/升格氛围;治不了「动作没演出来」——那是视频内容本身,要改 video_prompt 重出。有台词的镜慎用(人声也跟着变快变慢)'),
         image_prompt: z.string().optional().describe('首帧画面提示词**正文**(全量覆盖本镜现值)。★先 get_shot_prompts 读现值再改;★原样保留其中的 @char:N / @scene:M 引用标记,删了就不注入对应定妆图/场景图。出图时平台会在正文之上再拼身份锚与一致性约束,不必你写。★写法坑:别写「no X / without X / 不要 X / 没有 X」这类否定式约束——图像模型把名词当正向线索,反而把 X 画出来;要正向写出那块画面该有什么(材质/形状/颜色/远近)。保存响应带 image_prompt_negation_advisory 即命中,按其 note 改写'),
         video_prompt: z.string().optional().describe('视频(动态/运镜/表演)提示词**正文**(全量覆盖本镜现值)。★同 image_prompt:先读现值、保留 @char/@scene 标记'),
         first_frame_prompt: z.string().optional().describe('本镜**开始时**画面是什么样(首帧目标状态正文,全量覆盖现值)。送达提示词里作为 [START FRAME] 段排在最前,优先级高于 image_prompt——所以 image_prompt 改了不生效时,往往是这条在压着它。★同 image_prompt:先 get_shot_prompts 读现值、原样保留 @char/@scene 标记'),
@@ -1207,7 +1209,9 @@ export function registerProduceTools(server, client) {
         '★它也是「这一镜台词没念完/念了别的」最经济的修法:用克隆音重配整句并静音原声段,' +
         '**不重新生成视频**——按 TTS 档(千字符)计费,比 regenerate_shot_video(720p 212 点/秒)低几个数量级。' +
         '代价:画面口型是按原音演的,换音后可能对不上;口型看不清的镜(背身/远景/画外)几乎无损,大特写慎用。' +
-        '整集批量用 repair_episode_dialogue。', { storyboard_id: z.number().int().positive() }, async ({ storyboard_id }) => jsonResult(await client.producePost(`/storyboards/${storyboard_id}/dialogue-replace`)));
+        '★产物自动写回该镜(返回 persisted:true),但**进不进成片**取决于本集是否每镜都已合成:' +
+        '用 get_dialogue_repair_status 看 productEffective / pendingCompose,为 false 就先补合成再 rerender_episode,' +
+        '否则成片仍是旧音频。整集批量用 repair_episode_dialogue。', { storyboard_id: z.number().int().positive() }, async ({ storyboard_id }) => jsonResult(await client.producePost(`/storyboards/${storyboard_id}/dialogue-replace`)));
     server.tool('repair_episode_dialogue', '★整集批量修「台词没念完/念了别的」。默认只修 scan_dialogue_coverage 判为有缺陷的镜' +
         '(truncated/off_script/extra_speech);minor_gap 那一族是转写误差、原声本来是好的,不动。' +
         '原理与 replace_shot_dialogue 相同:克隆音重配整句 + 静音原声段,**不重新生成视频**——' +
@@ -1314,6 +1318,7 @@ export function registerProduceTools(server, client) {
     // ========== P1 · 整集配音(成片前的强制音频步)==========
     server.tool('assign_voices', '★给本集所有角色分配音色(配音导演 agent 按性别/性格/年龄/角色定位+项目语言选)。' +
         'voiceStyle 不是提取时自动填的——不分配,generate_tts 就没音色。文本步(LLM)后付。' +
+        '已用 set_character_voice 绑了客户自己授权音色的角色会被锁定、不会被改成别的音色。' +
         '后台异步,轮询 get_pipeline_status:assign_voices=done 即完成,再 generate_tts。', { episode_id: z.number().int().positive() }, async ({ episode_id }) => jsonResult(await client.producePost(`/episodes/${episode_id}/assign-voices`)));
     server.tool('generate_tts', '给整集所有对白批量配音(TTS)。成片前的音频步——不配音成片会缺对白。前置:先 assign_voices 给角色分配音色。后台异步,按用量后付不欠费。', { episode_id: z.number().int().positive() }, async ({ episode_id }) => jsonResult(await client.producePost(`/episodes/${episode_id}/tts`)));
     // ========== P1 · MV 子流程(project_type=mv 专用)==========
@@ -1506,7 +1511,17 @@ export function registerProduceTools(server, client) {
     server.tool('list_voices', '列出可用音色:我克隆的私有音色 + 平台公共音色(voice_id/名字/试听样本)。免费。', {}, async () => jsonResult(await client.produceGet('/voice-clones')));
     server.tool('delete_voice', '删除我音色库里的一个克隆音色(平台公共音色不可删)。免费。', { voice_id: z.string().describe('形如 lib:12') }, async ({ voice_id }) => jsonResult(await client.produceDelete(`/voice-clones/${String(voice_id).replace(/^lib:/, '')}`)));
     server.tool('set_character_voice', '把音色绑到某角色,之后 generate_tts 用它给该角色配音。免费。' +
-        '只有要出片时才需要这步——克隆和试听都不需要角色。', {
+        '只有要出片时才需要这步——克隆和试听都不需要角色。' +
+        '\n★★【原声剧(use_clip_audio=true,默认)里这一步就是「锁声线」】绑的是**客户自己的授权音色**' +
+        '(clone_voice 克隆的 / 客户上传的,不是公共音色)时,之后生成的视频会把这段声音直接交给视频厂商当参考音频,' +
+        '厂商按它发声、口型天然同步——否则每个组/单镜各自采样嗓音,同一角色前后镜声线会漂。' +
+        '公共音色只用于 TTS 配音,不会交给厂商。' +
+        '\n★代价:该角色出场的镜头生成视频时不再喂定妆视频锚(参考视频的音轨会压过参考音频,只能二选一),' +
+        '人物身份改由定妆图/设定图承担——所以绑之前先确认定妆图、设定图已经到位。' +
+        '\n★只对**绑定之后**生成的视频生效:要先绑、再 generate_videos。已经出好的视频要统一声线,' +
+        '用 repair_episode_dialogue(only_flagged=false,克隆音重配整句、不重生视频、按 TTS 档计费)最省;' +
+        '重生视频(regenerate_shot_video)也行但贵几个数量级。' +
+        '\n★绑好后 assign_voices 不会再覆盖这个角色(客户自己的声音优先级最高)。', {
         character_id: z.number().int().positive(),
         voice_id: z.string().describe('来自 clone_voice / list_voices 的 voice_id(形如 lib:12)'),
     }, async ({ character_id, voice_id }) => jsonResult(await client.producePost(`/characters/${character_id}/voice`, { voice_id })));

package/openapi.json

@@ -2,8 +2,8 @@
   "openapi": "3.1.0",
   "info": {
     "title": "StarReel Production API",
-    "version": "0.1.83",
-    "description": "Turn a script into a finished, downloadable short-drama episode over REST.\n\nPipeline: script → AI rewrite → cast/scenes/props extraction → storyboards → portraits & sheets + scene plates → keyframes → video shots → TTS → final cut (.mp4).\n\n**Billing is prepaid and agent-safe**: big-ticket stages are quote-then-generate (`quote_*` returns a `quote_id`). For video, quote == actual charge. For images the quote is a *range* — models billed per reference image are priced before those are assembled, so `estimated_points` is the upper bound (size the balance by it and a run never dies half-way) and `typical_points` is the usual cost; fixed-price models return the two equal. Insufficient balance returns 402 — nothing half-runs and the account never goes negative.\n\nAuth: exchange your API key at `POST /v1/agent/token` for a 15-minute bearer token.\n\nGenerated from the @starreel/mcp v0.1.83 tool surface (operationIds match MCP tool names 1:1)."
+    "version": "0.1.85",
+    "description": "Turn a script into a finished, downloadable short-drama episode over REST.\n\nPipeline: script → AI rewrite → cast/scenes/props extraction → storyboards → portraits & sheets + scene plates → keyframes → video shots → TTS → final cut (.mp4).\n\n**Billing is prepaid and agent-safe**: big-ticket stages are quote-then-generate (`quote_*` returns a `quote_id`). For video, quote == actual charge. For images the quote is a *range* — models billed per reference image are priced before those are assembled, so `estimated_points` is the upper bound (size the balance by it and a run never dies half-way) and `typical_points` is the usual cost; fixed-price models return the two equal. Insufficient balance returns 402 — nothing half-runs and the account never goes negative.\n\nAuth: exchange your API key at `POST /v1/agent/token` for a 15-minute bearer token.\n\nGenerated from the @starreel/mcp v0.1.85 tool surface (operationIds match MCP tool names 1:1)."
   },
   "servers": [
     {
@@ -493,7 +493,7 @@
       "post": {
         "operationId": "set_character_voice",
         "summary": "把音色绑到某角色,之后 generate_tts 用它给该角色配音",
-        "description": "把音色绑到某角色,之后 generate_tts 用它给该角色配音。免费。只有要出片时才需要这步——克隆和试听都不需要角色。",
+        "description": "把音色绑到某角色,之后 generate_tts 用它给该角色配音。免费。只有要出片时才需要这步——克隆和试听都不需要角色。\n★★【原声剧(use_clip_audio=true,默认)里这一步就是「锁声线」】绑的是**客户自己的授权音色**(clone_voice 克隆的 / 客户上传的,不是公共音色)时,之后生成的视频会把这段声音直接交给视频厂商当参考音频,厂商按它发声、口型天然同步——否则每个组/单镜各自采样嗓音,同一角色前后镜声线会漂。公共音色只用于 TTS 配音,不会交给厂商。\n★代价:该角色出场的镜头生成视频时不再喂定妆视频锚(参考视频的音轨会压过参考音频,只能二选一),人物身份改由定妆图/设定图承担——所以绑之前先确认定妆图、设定图已经到位。\n★只对**绑定之后**生成的视频生效:要先绑、再 generate_videos。已经出好的视频要统一声线,用 repair_episode_dialogue(only_flagged=false,克隆音重配整句、不重生视频、按 TTS 档计费)最省;重生视频(regenerate_shot_video)也行但贵几个数量级。\n★绑好后 assign_voices 不会再覆盖这个角色(客户自己的声音优先级最高)。",
         "tags": [
           "characters"
         ],
@@ -2835,7 +2835,7 @@
       "post": {
         "operationId": "assign_voices",
         "summary": "★给本集所有角色分配音色(配音导演 agent 按性别/性格/年龄/角色定位+项目语言选)",
-        "description": "★给本集所有角色分配音色(配音导演 agent 按性别/性格/年龄/角色定位+项目语言选)。voiceStyle 不是提取时自动填的——不分配,generate_tts 就没音色。文本步(LLM)后付。后台异步,轮询 get_pipeline_status:assign_voices=done 即完成,再 generate_tts。",
+        "description": "★给本集所有角色分配音色(配音导演 agent 按性别/性格/年龄/角色定位+项目语言选)。voiceStyle 不是提取时自动填的——不分配,generate_tts 就没音色。文本步(LLM)后付。已用 set_character_voice 绑了客户自己授权音色的角色会被锁定、不会被改成别的音色。后台异步,轮询 get_pipeline_status:assign_voices=done 即完成,再 generate_tts。",
         "tags": [
           "episodes"
         ],
@@ -8225,8 +8225,8 @@
     "/storyboards/{storyboard_id}": {
       "put": {
         "operationId": "update_shot",
-        "summary": "逐镜编辑:改单个分镜的文本内容(景别/动作/台词/画面描述/运镜等)与角色绑定(character_ids)",
-        "description": "逐镜编辑:改单个分镜的文本内容(景别/动作/台词/画面描述/运镜等)与角色绑定(character_ids)。只传要改的字段、其余不动。**免费**(纯文本写库)。★改 dialogue 会自动失效本镜已生成的 TTS 配音与字幕(需重出 tts);改文本不会自动重出图/视频,如需让画面跟上文本改动,改完再 regen 对应镜。用 get_storyboards 查改后结果。★★原声镜(厂商原生音频)改 dialogue 后,本镜视频会被标记「待重生」——因为台词是**烤进视频人声**的,不重生就终拼,成片里念的仍是改动前的台词(典型现象:台词像是跑到了别的镜头上)。compose_episode 会以 advisory `stale_video_after_edit` 列出这些镜;正确处置是先 regenerate_shot_video 再终拼。★也可直接改本镜的四条提示词正文(image_prompt/video_prompt/first_frame_prompt/last_frame_prompt):改前先用 get_shot_prompts 读现值,别凭空覆盖——正文里的 @char:N / @scene:M 是角色/场景参考图的引用标记,删掉本镜就不注入对应定妆图/场景图(形象漂移)。★★【出尾帧被拦下就是来改这里】出尾帧若报「此镜标为状态改变,但没有任何地方说明结束时是什么样子」(TERMINAL_DESC_GATE),修法是把**结束时画面是什么样**写进本镜 last_frame_prompt,然后重新出尾帧。别去重试出图——这类镜(action_motion_class=state_change 却没写终态)的尾帧历史成功率 9.4%,有终态描述的是 40.2%;不补描述就重生,送厂商的提示词里根本没有终态段,画出来仍和首帧一样、还是会被拒,而每次重生都照常扣费。get_shot_prompts 的 terminal_desc_missing=true 就是这个信号。",
+        "summary": "逐镜编辑:改单个分镜的文本内容(景别/动作/台词/画面描述/运镜等)、时长(duration)、成片变速(speed_factor)与角色绑定(character_ids)",
+        "description": "逐镜编辑:改单个分镜的文本内容(景别/动作/台词/画面描述/运镜等)、时长(duration)、成片变速(speed_factor)与角色绑定(character_ids)。只传要改的字段、其余不动。**免费**(纯文本写库)。★改 dialogue 会自动失效本镜已生成的 TTS 配音与字幕(需重出 tts);改文本不会自动重出图/视频,如需让画面跟上文本改动,改完再 regen 对应镜。用 get_storyboards 查改后结果。★★原声镜(厂商原生音频)改 dialogue 后,本镜视频会被标记「待重生」——因为台词是**烤进视频人声**的,不重生就终拼,成片里念的仍是改动前的台词(典型现象:台词像是跑到了别的镜头上)。compose_episode 会以 advisory `stale_video_after_edit` 列出这些镜;正确处置是先 regenerate_shot_video 再终拼。★也可直接改本镜的四条提示词正文(image_prompt/video_prompt/first_frame_prompt/last_frame_prompt):改前先用 get_shot_prompts 读现值,别凭空覆盖——正文里的 @char:N / @scene:M 是角色/场景参考图的引用标记,删掉本镜就不注入对应定妆图/场景图(形象漂移)。★★【出尾帧被拦下就是来改这里】出尾帧若报「此镜标为状态改变,但没有任何地方说明结束时是什么样子」(TERMINAL_DESC_GATE),修法是把**结束时画面是什么样**写进本镜 last_frame_prompt,然后重新出尾帧。别去重试出图——这类镜(action_motion_class=state_change 却没写终态)的尾帧历史成功率 9.4%,有终态描述的是 40.2%;不补描述就重生,送厂商的提示词里根本没有终态段,画出来仍和首帧一样、还是会被拒,而每次重生都照常扣费。get_shot_prompts 的 terminal_desc_missing=true 就是这个信号。",
         "tags": [
           "storyboards"
         ],
@@ -8291,6 +8291,14 @@
                   "shot_intent": {
                     "type": "string"
                   },
+                  "duration": {
+                    "type": "number",
+                    "exclusiveMinimum": 0
+                  },
+                  "speed_factor": {
+                    "type": "number",
+                    "exclusiveMinimum": 0
+                  },
                   "image_prompt": {
                     "type": "string"
                   },
@@ -8340,7 +8348,7 @@
       "post": {
         "operationId": "replace_shot_dialogue",
         "summary": "换某镜对白音色/声线(转写+克隆重配)",
-        "description": "换某镜对白音色/声线(转写+克隆重配)。后台异步,按用量后付不欠费。要求本镜有原声视频+角色声线定妆音。★它也是「这一镜台词没念完/念了别的」最经济的修法:用克隆音重配整句并静音原声段,**不重新生成视频**——按 TTS 档(千字符)计费,比 regenerate_shot_video(720p 212 点/秒)低几个数量级。代价:画面口型是按原音演的,换音后可能对不上;口型看不清的镜(背身/远景/画外)几乎无损,大特写慎用。整集批量用 repair_episode_dialogue。",
+        "description": "换某镜对白音色/声线(转写+克隆重配)。后台异步,按用量后付不欠费。要求本镜有原声视频+角色声线定妆音。★它也是「这一镜台词没念完/念了别的」最经济的修法:用克隆音重配整句并静音原声段,**不重新生成视频**——按 TTS 档(千字符)计费,比 regenerate_shot_video(720p 212 点/秒)低几个数量级。代价:画面口型是按原音演的,换音后可能对不上;口型看不清的镜(背身/远景/画外)几乎无损,大特写慎用。★产物自动写回该镜(返回 persisted:true),但**进不进成片**取决于本集是否每镜都已合成:用 get_dialogue_repair_status 看 productEffective / pendingCompose,为 false 就先补合成再 rerender_episode,否则成片仍是旧音频。整集批量用 repair_episode_dialogue。",
         "tags": [
           "storyboards"
         ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@starreel/mcp",
-  "version": "0.1.83",
+  "version": "0.1.85",
   "mcpName": "ai.starreel/starreel",
   "description": "StarReel MCP server — drive the AI short-drama production pipeline (script → storyboards → frames → video → final cut) from Claude, Cursor, or any MCP client",
   "license": "MIT",

package/server.json

@@ -2,13 +2,13 @@
   "$schema": "https://static.modelcontextprotocol.io/schemas/2025-09-29/server.schema.json",
   "name": "ai.starreel/starreel",
   "description": "Turn a script into a finished, downloadable short-drama episode — via MCP or REST.",
-  "version": "0.1.83",
+  "version": "0.1.85",
   "websiteUrl": "https://starreel.ai",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "@starreel/mcp",
-      "version": "0.1.83",
+      "version": "0.1.85",
       "transport": {
         "type": "stdio"
       },