zerocut-cli 0.4.0 → 0.4.1

package/dist/skill/one-click-video/SKILL.md

@@ -34,7 +34,7 @@ Based on user topic, purpose, duration, style, references, and subject assets:
 
 Required workflow:
 
-project preparation -> subject creation -> scene planning (ensure `./materials/scene-bible.md` exists) -> storyboard generation -> scene video generation -> BGM generation -> final composition
+project preparation -> subject creation -> scene planning (ensure `./scene-bible.md` exists) -> storyboard generation -> scene video generation -> BGM generation -> final composition
 
 ## Runtime Preparation
 
@@ -89,6 +89,10 @@ npx zerocut-cli config --ott <token> --region <cn|us>
 
 - If user does not provide exact script, generate concise narration/dialogue aligned with story and duration.
 - If user provides script or key message, keep original intent, wording priority, and brand keywords.
+- If narration exists for a scene, inject it at the beginning of the scene video prompt with this exact format:
+  - `【narration_tone】Narration:<text_content>\n`
+- `narration_tone` must follow the voice design rules in this skill.
+- `text_content` must be written in the user-required language.
 - Estimate speech duration with normal-slow pace.
 - Per-scene narration/dialogue total should not exceed 12 seconds.
 - If over limit, compress script or split into more scenes.
@@ -115,7 +119,7 @@ npx zerocut-cli image --prompt "<subject turnaround prompt>" --model banana2 --t
 
 ## Scene Planning And Scene-Bible Rules
 
-Scene planning is a critical quality gate. Before any storyboard or video generation, create a complete `./materials/scene-bible.md` and treat it as the single source of truth for all downstream prompts.
+Scene planning is a critical quality gate. Before any storyboard or video generation, create a complete `./scene-bible.md` and treat it as the single source of truth for all downstream prompts.
 
 ### Mandatory Scene-Bible Checklist
 
@@ -136,7 +140,7 @@ Scene planning is a critical quality gate. Before any storyboard or video genera
 7. **Shot plan per scene**
    - each shot has self-contained prompt requirements and camera intention
 8. **Narration/dialogue plan**
-   - per scene script and estimated speech duration (must stay within 12s per scene)
+   - per scene script, narration tone, language, and estimated speech duration (must stay within 12s per scene)
 9. **Asset binding**
    - which references are required for each scene (`--storyboard`, `--persons`, `--refs`)
 10. **Output plan**
@@ -146,7 +150,7 @@ Scene planning is a critical quality gate. Before any storyboard or video genera
 
 ### Scene-Bible Template
 
-Use this structure when writing `./materials/scene-bible.md`:
+Use this structure when writing `./scene-bible.md`:
 
 ```markdown
 # Scene Bible
@@ -280,6 +284,8 @@ For each scene, planning output must include:
 ## Storyboard Rules
 
 - Every scene must have a storyboard image.
+- Storyboard generation must use `npx zerocut-cli image` with `--type storyboard`.
+- `--type` value is strictly locked to `storyboard` for storyboard generation. Do not use `default`, `subject-turnaround`, or any other value.
 - Storyboard must include environment, subject position/action, framing, camera movement, rhythm, and key emotion.
 - If a subject appears in that scene, include matching subject references.
 - Storyboard prompts must be complete and self-contained.
@@ -295,6 +301,8 @@ If no subject references are required in that scene, omit `--refs`.
 ## Prompt Independence Hard Constraint
 
 - Every shot prompt and every scene video prompt must be fully self-contained.
+- If narration is present, the prompt must start with `【narration_tone】Narration:<text_content>\n`.
+- Validate narration before generation: tone must match voice design and text language must match user requirement.
 - Do not use shorthand such as:
   - "same as previous shot"
   - "continue above"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "zerocut-cli",
-  "version": "0.4.0",
+  "version": "0.4.1",
   "description": "ZeroCut CLI: AI assistant CLI for creating and editing images/audio/video",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

package/src/skill/one-click-video/SKILL.md

@@ -34,7 +34,7 @@ Based on user topic, purpose, duration, style, references, and subject assets:
 
 Required workflow:
 
-project preparation -> subject creation -> scene planning (ensure `./materials/scene-bible.md` exists) -> storyboard generation -> scene video generation -> BGM generation -> final composition
+project preparation -> subject creation -> scene planning (ensure `./scene-bible.md` exists) -> storyboard generation -> scene video generation -> BGM generation -> final composition
 
 ## Runtime Preparation
 
@@ -89,6 +89,10 @@ npx zerocut-cli config --ott <token> --region <cn|us>
 
 - If user does not provide exact script, generate concise narration/dialogue aligned with story and duration.
 - If user provides script or key message, keep original intent, wording priority, and brand keywords.
+- If narration exists for a scene, inject it at the beginning of the scene video prompt with this exact format:
+  - `【narration_tone】Narration:<text_content>\n`
+- `narration_tone` must follow the voice design rules in this skill.
+- `text_content` must be written in the user-required language.
 - Estimate speech duration with normal-slow pace.
 - Per-scene narration/dialogue total should not exceed 12 seconds.
 - If over limit, compress script or split into more scenes.
@@ -115,7 +119,7 @@ npx zerocut-cli image --prompt "<subject turnaround prompt>" --model banana2 --t
 
 ## Scene Planning And Scene-Bible Rules
 
-Scene planning is a critical quality gate. Before any storyboard or video generation, create a complete `./materials/scene-bible.md` and treat it as the single source of truth for all downstream prompts.
+Scene planning is a critical quality gate. Before any storyboard or video generation, create a complete `./scene-bible.md` and treat it as the single source of truth for all downstream prompts.
 
 ### Mandatory Scene-Bible Checklist
 
@@ -136,7 +140,7 @@ Scene planning is a critical quality gate. Before any storyboard or video genera
 7. **Shot plan per scene**
    - each shot has self-contained prompt requirements and camera intention
 8. **Narration/dialogue plan**
-   - per scene script and estimated speech duration (must stay within 12s per scene)
+   - per scene script, narration tone, language, and estimated speech duration (must stay within 12s per scene)
 9. **Asset binding**
    - which references are required for each scene (`--storyboard`, `--persons`, `--refs`)
 10. **Output plan**
@@ -146,7 +150,7 @@ Scene planning is a critical quality gate. Before any storyboard or video genera
 
 ### Scene-Bible Template
 
-Use this structure when writing `./materials/scene-bible.md`:
+Use this structure when writing `./scene-bible.md`:
 
 ```markdown
 # Scene Bible
@@ -280,6 +284,8 @@ For each scene, planning output must include:
 ## Storyboard Rules
 
 - Every scene must have a storyboard image.
+- Storyboard generation must use `npx zerocut-cli image` with `--type storyboard`.
+- `--type` value is strictly locked to `storyboard` for storyboard generation. Do not use `default`, `subject-turnaround`, or any other value.
 - Storyboard must include environment, subject position/action, framing, camera movement, rhythm, and key emotion.
 - If a subject appears in that scene, include matching subject references.
 - Storyboard prompts must be complete and self-contained.
@@ -295,6 +301,8 @@ If no subject references are required in that scene, omit `--refs`.
 ## Prompt Independence Hard Constraint
 
 - Every shot prompt and every scene video prompt must be fully self-contained.
+- If narration is present, the prompt must start with `【narration_tone】Narration:<text_content>\n`.
+- Validate narration before generation: tone must match voice design and text language must match user requirement.
 - Do not use shorthand such as:
   - "same as previous shot"
   - "continue above"