@aswless_854771076/ai_short_studio_cli 0.1.152 → 0.1.153
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/package.json +1 -1
- package/skills/using-vvicat-ai-short-studio-cli/SKILL.md +1 -1
- package/skills/using-vvicat-ai-short-studio-cli/references/project-and-canvas.md +13 -1
- package/skills/using-vvicat-ai-short-studio-cli/references/short-drama/assets/modules/design-drama-audio/prompts/voice_analysis.en.txt +3 -2
- package/skills/using-vvicat-ai-short-studio-cli/references/short-drama/assets/modules/design-drama-audio/prompts/voice_analysis.zh.txt +3 -2
- package/skills/using-vvicat-ai-short-studio-cli/references/short-drama/assets/modules/plan-drama-production/prompts/production_plan.en.txt +1 -1
- package/skills/using-vvicat-ai-short-studio-cli/references/short-drama/generation/design-drama-audio.md +1 -1
package/package.json
CHANGED
|
@@ -56,7 +56,7 @@ description: Use when an agent needs to inspect or operate VVICAT infinite-canva
|
|
|
56
56
|
|
|
57
57
|
1. **预检与恢复**:运行 `preflight --json`;更新后重跑。再读取 `config get`、目标项目、`canvas documents` 和 `canvas workflow status`。只有历史项目需要新增媒体节点时才读取实时 `canvas node types`。`skill status --target codex` 非 current 时先询问再更新,禁止强制覆盖用户修改。
|
|
58
58
|
2. **项目语言门禁**:新建项目前必须从用户确认交付语言,并用 `project create --language <BCP47>` 写入;界面 `--locale` 不能代替内容语言。创建后回读 `project get` 与 `canvas settings get`,确认两者语言一致且不可修改。项目语言约束剧本、分镜、台词、配音、画面文字和字幕;验收须区分剧情内容与制作信息:项目语言允许正常外语穿插和规范英文摄影术语,不按字符种类判死;道具文字和剧情内界面遵循来源原文。存在实际语言合同冲突时停止并说明。
|
|
59
|
-
3.
|
|
59
|
+
3. **项目配置门禁**:项目初始化时必须向用户展示并确认创作配置,不得只采用账户默认值:交付语言、画幅、目标时长、画风、分镜板策略、分析/文本模型、图片模型及其厂商、视频模型及其厂商、视频提示词协议、音频/配乐模型及其厂商、声音一致性方式、分辨率、画质、候选数量、参考素材和预计请求范围。先读取 `canvas settings get`,展示 `currentValue`、`effectiveValue`、`source`、建议值和实时 options;确认 `imageResolution`、`imageQuality` 和 `videoResolution` 后再用 `--file settings.json` 原子写入并携带 `expectedVersion`,回读一致前不得创建或运行付费节点。`$ART_STYLE_ID` 必须解析实时 `option.value`。节点确需不同参数时才按实时能力覆盖,并记录覆盖原因;`automation_mode` 只能省略已确认且未变化的重复询问,不能替用户确认模型/厂商或付费范围。
|
|
60
60
|
4. **自由画布文档和目标集**:先读取 `canvas workflow status`,按当前阶段和条件模块生成 JSON;使用 `canvas document save` 写入 `hypit.<module>.v1`,用 `--input` 固化上游选版。审核必须另存为 `hypit.review.v1` 并绑定被审核版本。无需先调用固定节点或生成模型,不要创建没有版本的文档空壳。已有自由画布文档直接复用;历史 `screenplay-resource` 或结构化资源先迁移成 自由画布文档,再继续流程。需要创作微短剧时,按需读取 `references/short-drama/`,再读取 `references/humanizer/` 去除 AI 味;不得改动已经确认的剧情事实,润色后回到短剧 reference 复核。确认目标集数后,只生产该集实际需要的资产。
|
|
61
61
|
5. **资产门禁**:先完成人物、场景、道具及其派生版本。每批生成完成后必须用 `canvas asset download` 或等效预览能力取得真实候选,并在当前对话中展示候选缩略图/文件链接、资产 key、版本和用途;用户确认后才能写入 selected version。只在用户看过候选且确认后选版。执行资产—剧情匹配审计;缺失、多余、错版、stale、错边未清零时拆镜前门禁失败。
|
|
62
62
|
6. **拆镜和滚动生成**:先执行 [拆镜模式与交付范围](references/assets-and-storyboard.md#拆镜模式与交付范围) 和 [语言匹配门禁](references/assets-and-storyboard.md#拆镜语言匹配门禁)。短剧新建默认专业拆镜,显式设置 `sourceFormat` 与 `storyboardPipelineVersion: 2`,不自动新增首帧节点。Agent 先读取每镜 `storyboard_strategy`,按制作目标选择单幅分镜图、时间故事版、镜头分镜板、专项板或 Blender;已选择一种板型时不得自动补生成另一种。只审核实际生成的板型,并让视频提示词只引用真实存在且已审核的分镜资产。用户的停止阶段优先于 [滚动异步调度](references/async-pipeline.md):仅分镜图任务完成实际选择的板型并通过审核后停止,不运行视频提示词和视频。允许视频时,已就绪镜头滚动提交,单镜失败不阻塞其他镜头。
|
|
@@ -4,13 +4,25 @@
|
|
|
4
4
|
|
|
5
5
|
## 初始化和配置
|
|
6
6
|
|
|
7
|
-
1. `preflight --json` 检查地址、CLI 版本、认证和标准能力;`restartRequired`
|
|
7
|
+
1. `preflight --json` 检查地址、CLI 版本、认证和标准能力;`restartRequired` 为真时重跑。项目初始化先向用户确认交付语言、画幅、目标时长、画风、分镜策略、各模态模型及厂商、提示词协议、声音一致性方式、分辨率、画质、候选数量和请求范围;不得把账户默认模型或厂商静默当成用户选择。仅在用户已确认且未发生变化时,缺少可确定的账户 `capabilityDefaults` 才能从实时 options、目标项目有效值或模型默认值补齐并回读;模型不可用或候选缺失时停下询问。
|
|
8
8
|
2. `config get` 只管理个人默认模型、能力默认值和工作流并发。供应商、密钥和全局模型目录由管理员维护。
|
|
9
9
|
3. 新项目按“确认名称/用途与交付语言 → `project create --language <BCP47>` → `project get` + 一次 `canvas inspect`”执行;既有项目先聚合读取。项目配置回读通过前不得创建创作节点。
|
|
10
10
|
4. `canvas settings get` 必须确认项目覆盖值、继承后的有效值、来源和实时候选。默认推荐 `storyboardImageType: grid`;`visualBible` 按实时 schema 构造。音频配音与音乐生成使用独立槽位:`audioModel` 只服务 TTS,`musicModel` 只服务 `music-generate`;不要用 TTS 模型填充音乐槽位。项目 `outputLanguage` 必须与创建时用户确认的交付语言一致;它在创建后不可修改,画布设置只读返回,不能使用 `null` 或清空。CLI `--locale` 只影响命令界面和输出消息,不代表内容语言。
|
|
11
11
|
5. `imageResolution`、`imageQuality`、`videoResolution` 优先由项目统一设置;节点只有确需不同且实时能力允许时才覆盖。H3 模型必须与 `videoPromptType: "h3"` 同次写入。
|
|
12
12
|
6. 用一次 `canvas settings set --file` 原子写入全部确认项并携带最近的 `expectedVersion`。409 后重新读取、展示变化并确认;写后立即回读。
|
|
13
13
|
|
|
14
|
+
### 声音一致性绑定
|
|
15
|
+
|
|
16
|
+
初始化时单独询问每个需要说话的人物的声音来源,并把选择写入 `audio-plan.voice_consistency`:
|
|
17
|
+
|
|
18
|
+
- `first_appearance_sample`(默认):找到人物第一次出场且有清晰对白的已选视频或原声,截取实际说话区间作为声纹锚点;先展示截取时长、来源镜头、波形/试听链接和权利状态,用户确认后才能上传或绑定。
|
|
19
|
+
- `user_upload`:用户上传并确认有权使用的声音样本。
|
|
20
|
+
- `global_asset`:从全局声音资产库选择已审核的 voice ID 和版本。
|
|
21
|
+
- `provider_voice_id`:选择实时 Provider 目录中已确认的声音 ID。
|
|
22
|
+
- `none`:明确不做跨镜声音一致性绑定,后续每镜按项目确认的声音策略生成。
|
|
23
|
+
|
|
24
|
+
默认方案只负责建立绑定,不自动克隆或发起付费生成;绑定记录必须包含 `character_id`、`source_mode`、来源资产/镜头、起止时间、SHA-256、试听证据和用户确认记录。后续独立配音、电话音轨或视频原生声音只能引用已确认的绑定,绑定变更必须重新试听并复审。
|
|
25
|
+
|
|
14
26
|
## 查询与批处理
|
|
15
27
|
|
|
16
28
|
- 常规查询先用一次 `canvas inspect`,通过 `--include`、`--kind`、`--status`、`--search`、`--operation`、`--resource-type`、`--producer-node` 表达;现有命令不能表达时才写临时脚本。
|
|
@@ -30,7 +30,8 @@ Rules:
|
|
|
30
30
|
7. New plans must set `audio_strategy.mode=native-first`, `provider_selection=prefer-native`, and `fallback_allowed=true`. `fallback_reasons` must contain exactly: `provider-no-native-audio`, `voice-identity-drift`, `speech-intelligibility-failed`, `narration-performance-failed`, `audio-sync-failed`, and `native-ambience-failed`.
|
|
31
31
|
8. Keep source and presentation orthogonal on every line: `delivery_mode=native|post_dub|external_audio`, `presentation=visible-dialogue|offscreen-dialogue|narration`, and `fallback_mode=none|post-dub|cinematic-tts|sound-design`. Never put narration or offscreen in delivery_mode.
|
|
32
32
|
9. Use null for `source_audio` and `voice_binding` when no selected asset or real voice evidence exists. A native-audio fallback requires `native_audio_exception` with an allowed reason, actual evidence, an exact millisecond range, and final `mix_sources`. Offscreen dialogue or narration that was never generated is not a native-audio failure and must keep `native_audio_exception` null.
|
|
33
|
-
9a.
|
|
33
|
+
9a. `voice_consistency.source_mode` must be one of `first_appearance_sample|user_upload|global_asset|provider_voice_id|none`, taken from the user-confirmed project initialization. `first_appearance_sample` extracts a real interval from the selected video or source audio where the character first appears with clear dialogue, and records `source_asset_key`, `source_version_id`, `start_ms`, `end_ms`, `sha256`, preview evidence, and confirmation; without preview and confirmation put it in unresolved and never bind automatically. `user_upload` needs an upload receipt and rights confirmation, `global_asset` needs a reviewed global voice ID/version, and `provider_voice_id` needs live catalog evidence; `none` keeps voice_binding null.
|
|
34
|
+
9b. Missing offscreen dialogue uses `delivery_mode=post_dub`, `presentation=offscreen-dialogue`, and a clone of the character's original voice made with the model required by the current audio specification. Its voice binding must include `voice_role=character`, `source=clone`, the real `target_model`, and a real `upload_receipt_id`. Narration uses `delivery_mode=post_dub`, `presentation=narration`, and an independently designed professional narrator voice. Its binding must include `voice_role=narrator`, `source=design`, `professional_narration=true`, a real `request_id`, and a `cinematic_profile` identical to performance. Both routes must bind source `speech-timing`; never estimate timing from text.
|
|
34
35
|
10. Every narration line requires an executable cinematic `performance` with `tone_arc`, ordered `emotion_beats[]`, `pace`, `breath_and_pause`, and `distance_and_space`; “emotional” alone is insufficient. Non-narration performance is null. Native background sound means ambience and action sound only; BGM remains an independent licensed-music workflow.
|
|
35
36
|
|
|
36
|
-
Return one strict JSON object only: {"episode_key":"...","source_versions":{},"audio_strategy":{"mode":"native-first","provider_selection":"prefer-native","fallback_allowed":true,"fallback_reasons":[]},"lines":[],"voice_bindings":[],"unresolved":[],"approved":false}. Every lines item must contain exactly: line_index, speaker, language, line_type, content, emotion, emotion_strength, pronunciation_notes, matched_shot, delivery_mode, presentation, fallback_mode, source_audio, voice_binding, native_audio_exception, performance. line_index starts at 1 and is consecutive; pronunciation_notes is a string array; matched_shot is {"shot_number": positive integer} or null. Use an empty lines array when nothing must be heard. voice_bindings records only tool-receipted, user-confirmed speaker, provider, model, target_model, voice_id, voice_role, and
|
|
37
|
+
Return one strict JSON object only: {"episode_key":"...","source_versions":{},"audio_strategy":{"mode":"native-first","provider_selection":"prefer-native","fallback_allowed":true,"fallback_reasons":[]},"lines":[],"voice_bindings":[],"unresolved":[],"approved":false}. Every lines item must contain exactly: line_index, speaker, language, line_type, content, emotion, emotion_strength, pronunciation_notes, matched_shot, delivery_mode, presentation, fallback_mode, source_audio, voice_binding, native_audio_exception, performance. line_index starts at 1 and is consecutive; pronunciation_notes is a string array; matched_shot is {"shot_number": positive integer} or null. Use an empty lines array when nothing must be heard. voice_bindings records only tool-receipted, user-confirmed speaker, provider, model, target_model, voice_id, voice_role, source, and source_mode; `first_appearance_sample` adds source_asset_key/source_version_id/start_ms/end_ms/sha256, `user_upload` adds upload_receipt_id, `global_asset` adds global_asset_id/global_asset_version, `provider_voice_id` adds catalog_evidence, cloned voices add upload_receipt_id, designed voices add request_id, and professional narration adds professional_narration plus cinematic_profile. Put unresolved source, speaker, language, pronunciation, performance, or parameter evidence in unresolved and keep approved false.
|
|
@@ -30,7 +30,8 @@
|
|
|
30
30
|
7. 新计划固定 `audio_strategy.mode=native-first`、`provider_selection=prefer-native`、`fallback_allowed=true`;`fallback_reasons` 必须完整且仅含 `provider-no-native-audio`、`voice-identity-drift`、`speech-intelligibility-failed`、`narration-performance-failed`、`audio-sync-failed`、`native-ambience-failed`。
|
|
31
31
|
8. 每行分别填写声音来源 `delivery_mode=native|post_dub|external_audio`、画面呈现 `presentation=visible-dialogue|offscreen-dialogue|narration`、预案 `fallback_mode=none|post-dub|cinematic-tts|sound-design`,不得把 narration/offscreen 填进 delivery_mode。
|
|
32
32
|
9. `source_audio` 和 `voice_binding` 没有已选资产或真实音色证据时为 null,不得伪造。原生失败兜底时 `native_audio_exception` 必须写受控 reason、实际 evidence、毫秒 range 和最终 `mix_sources`;原本未生成的场外音或旁白不是原生失败,`native_audio_exception` 必须为 null。
|
|
33
|
-
9a.
|
|
33
|
+
9a. `voice_consistency.source_mode` 只能是 `first_appearance_sample|user_upload|global_asset|provider_voice_id|none`,且必须来自项目初始化时用户确认的配置。`first_appearance_sample` 从人物第一次出场且有清晰对白的已选视频或原声截取真实区间,写入 `source_asset_key`、`source_version_id`、`start_ms`、`end_ms`、`sha256`、试听证据和确认记录;未展示试听或未确认时写入 unresolved,不得自动绑定。`user_upload` 必须有上传回执和权利确认,`global_asset` 必须有全局 voice ID/版本和审核证据,`provider_voice_id` 必须有实时目录证据;`none` 时 voice_binding 为 null。
|
|
34
|
+
9b. 原本未生成的场外音设 `delivery_mode=post_dub`、`presentation=offscreen-dialogue`,绑定当前声音规范指定模型克隆的角色原声音色;voice_bindings 必须含 `voice_role=character`、`source=clone`、真实 `target_model` 和 `upload_receipt_id`。旁白设 `delivery_mode=post_dub`、`presentation=narration`,使用独立专业设计音色;voice_bindings 必须含 `voice_role=narrator`、`source=design`、`professional_narration=true`、真实 `request_id`,且 `cinematic_profile` 与 performance 逐字段一致。两者都必须先绑定原声 speech-timing,不得按文字估算时间。
|
|
34
35
|
10. `presentation=narration` 必须写电影感 `performance`:`tone_arc`、按顺序的 `emotion_beats[]`、`pace`、`breath_and_pause`、`distance_and_space`,不能只写“富有情感”;非旁白为 null。背景音原生直出只包括环境声和动作声,BGM 仍由独立授权配乐流程负责。
|
|
35
36
|
|
|
36
|
-
只返回严格 JSON 对象:{"episode_key":"...","source_versions":{},"audio_strategy":{"mode":"native-first","provider_selection":"prefer-native","fallback_allowed":true,"fallback_reasons":[]},"lines":[],"voice_bindings":[],"unresolved":[],"approved":false}。lines 每项必须且只能包含:line_index、speaker、language、line_type、content、emotion、emotion_strength、pronunciation_notes、matched_shot、delivery_mode、presentation、fallback_mode、source_audio、voice_binding、native_audio_exception、performance。line_index 从 1 连续递增;pronunciation_notes 为字符串数组;matched_shot 为 {"shot_number": 正整数} 或 null。无可听文本时 lines 返回 []。voice_bindings 只记录用户已确认且有工具收据的 speaker、provider、model、target_model、voice_id、voice_role、source
|
|
37
|
+
只返回严格 JSON 对象:{"episode_key":"...","source_versions":{},"audio_strategy":{"mode":"native-first","provider_selection":"prefer-native","fallback_allowed":true,"fallback_reasons":[]},"lines":[],"voice_bindings":[],"unresolved":[],"approved":false}。lines 每项必须且只能包含:line_index、speaker、language、line_type、content、emotion、emotion_strength、pronunciation_notes、matched_shot、delivery_mode、presentation、fallback_mode、source_audio、voice_binding、native_audio_exception、performance。line_index 从 1 连续递增;pronunciation_notes 为字符串数组;matched_shot 为 {"shot_number": 正整数} 或 null。无可听文本时 lines 返回 []。voice_bindings 只记录用户已确认且有工具收据的 speaker、provider、model、target_model、voice_id、voice_role、source、source_mode;`first_appearance_sample` 追加 source_asset_key/source_version_id/start_ms/end_ms/sha256,`user_upload` 追加 upload_receipt_id,`global_asset` 追加 global_asset_id/global_asset_version,`provider_voice_id` 追加 catalog_evidence,克隆音色追加 upload_receipt_id,设计音色追加 request_id,专业旁白追加 professional_narration 与 cinematic_profile。任何来源、说话人、语言、读音、表演或参数未确认时写入 unresolved,approved 必须为 false。
|
|
@@ -36,7 +36,7 @@ Rules:
|
|
|
36
36
|
7. prompt_profile is exactly seedance2 or h3. Final generation uses character images, a temporal storyboard, a shot board, an audio reference, and the video prompt. Add Blender previz only when the two still boards cannot prove a complex route, multi-person interaction, exact contact, axis risk, or continuous camera move. Seedance 2.0 uses `full-reference`, and H3 uses `Ref2VA`. If the Provider cannot carry the inputs this shot actually requires, mark unresolved; never silently omit one.
|
|
37
37
|
8. When the Provider supports MiniMax H3 native audio (video.native-audio), default audio_strategy.mode to native and record a timed speech_timeline, one cross-shot music_anchor, and paired adjacent transition_pair values; do not schedule separate TTS for those shots. Use post-dub/independent only when the user explicitly requests dubbing or the Provider/model is confirmed to lack native audio; do not silently fall back to dubbing just because capability is unknown. Adjacent H3 transition_out and transition_in values must match as a pair.
|
|
38
38
|
9. Every shot uses image_strategy.mode=generate, board_type=shot-board, and panel_grid_size=2..16 (default 6, adjusted to actual beats), producing independent temporal-storyboard and shot-board assets. Standard projects default to storyboard_strategy.mode=image and previz_strategy={"mode":"none"}. Use storyboard_strategy.mode=blender with previz_strategy={"mode":"blender","purpose":"motion-reference","fps":12} only when the two boards still cannot unambiguously express a complex route, multi-person interaction, exact contact, axis risk, or continuous camera move; name that spatial risk in review_checks. Only StarRouter MiniMax-H3 currently accepts the Blender motion reference through TempFile.org hosting; otherwise mark unresolved. Do not bind future asset versions. Composition elements require source, selected-asset, or confirmed-blocking evidence; block unsupported elements in unresolved. overflow_strategy is exactly compose-assets or reject.
|
|
39
|
-
9a. Before generation, realize `required_board_types`: a video shot must generate
|
|
39
|
+
9a. Before generation, realize `required_board_types`: a video shot must generate only the board types selected in its `storyboard_strategy.required_board_types`; narrative, blocking, action, choreography, comprehensive, director-track, temporal, shot-board, and previz-3d are selectable types. Add multiple types only when the production goal needs each one; `previz-3d` is an additional spatial reference when selected. Do not auto-create an unselected board type. `image_strategy.board_type=shot-board` is only a compatibility field for the executor and must not override `primary_board_type`.
|
|
40
40
|
10. Tailor review_checks to actual shot risk; every item must be a concrete, named, observable phrase (e.g. "mirror reflection consistency", "contact point readable through rain") derived from the contract's ai_risks and evidence elements: dialogue checks speaker identity and every visible mouth state; group shots check headcount, order, facing, and limb ownership; prop interaction checks before/contact/after states and adjacent objects; walking or fights check foot contact, weight, direction, speed, and pose across cuts; in-frame screens check whether precise post-production graphics are required. When composition risks apply, turn each into an explicit check: mirrors/glossy water verify the reflection matches the real person and actual dressing, with no extra fingers, camera, or reflection anomaly; rain, fog, neon, and smoke/dust verify the weather and colored light are evidenced, directions are consistent, and neither buries the contact point nor identity; crowd shots verify headcount, identity, costume-color anchors, and no extra bystanders; weapons and multi-hand interaction verify prop form, holding relationship, hand count, and before/contact/after states; a narrow door gap or foreground occlusion verifies visible width, both eyes intact, and the key action uncovered; high-speed trails verify both eyes, the face, and the contact point stay sharp and the subject is not a blob; front-and-back pincer/encirclement verifies both threats' identity, distance, and facing stay readable in foreground and background; vertical shots verify faces, key props, and contact points sit in the upper-middle safe area, the bottom subtitle safe area is clear, and motion edges keep a 10% margin. Keep only the risks this shot actually has. Never copy one generic checklist onto every shot.
|
|
41
41
|
|
|
42
42
|
11. Only a shot whose previz_strategy.mode is blender runs direct-blender-previz before generate-blender-previz renders and scores it. Previz never replaces the temporal storyboard, shot board, character, scene, prop, wardrobe, material, or final-look references.
|
|
@@ -2,7 +2,7 @@
|
|
|
2
2
|
|
|
3
3
|
优先在视频模型支持时保留原生对白、环境声和动作声。逐镜审核语言、说话者、文本、可懂度、时间同步、环境动作声和未声明音乐。失败时记录实际区间与原因;合格部分不重复生成。
|
|
4
4
|
|
|
5
|
-
独立语音先用 `voice_analysis`
|
|
5
|
+
独立语音先用 `voice_analysis` 分离说话人与台词,再按项目初始化时确认的 `audio-plan.voice_consistency` 绑定声音。默认 `first_appearance_sample` 从人物第一次出场且有清晰对白的已选视频或原声截取实际说话区间,保存来源镜头、起止时间、SHA-256 并先展示试听;用户确认后才可作为 voice reference。`user_upload` 使用用户上传且已确认有权使用的样本,`global_asset` 使用全局资产库已审核 voice ID,`provider_voice_id` 使用实时 Provider 目录声音,`none` 不建立跨镜绑定。需要复用原声时,才使用 `voice-clone` 生成真实音色 ID;需要新声音时使用 `voice-design`。生成前核对逐字文本和使用权,输出作为未选候选逐条试听、审核后再选版。普通语音模型可用于已确认参数的配音;模型可用性由实时目录和运行时探测。已授权的外部音频可以导入项目并按同样流程审核。
|
|
6
6
|
|
|
7
7
|
CosyVoice 自定义音色设计/克隆在独立 TTS 流程中可用,但必须使用与目标模型匹配的真实音色 ID,并经过试听和审核;它不能覆盖 H3 原生对白。MuseTalk 与旧 RunningHub 口型驱动仍暂时禁用。旧 audio-plan 中的失效配置不得继续提交付费任务;向用户说明缺失能力,并先重新确认当前可用模型和制作方案。不得把临时音色或未通过审核的配音默认为正式结果。
|
|
8
8
|
|