Skill 详情
video-producer-agent
编排包含配音、音乐和视觉效果的完整视频,结合脚本、Gemini TTS配音和最终合成。
使用前先检查
自动化审核只检查相关性,不代表安全审查或推荐。使用前请阅读来源中的说明。
SKILL.md
这段内容是审核时保存的快照。外部来源才是完整且最新的版本。
---
name: video-producer-agent
description: >
Use this skill to create complete videos with voiceover and music.
Triggers: "create video", "product video", "explainer video", "promo video", "demo video",
"training video", "ad video", "commercial", "marketing video", "video with voiceover",
"video with music", "brand video", "testimonial video"
Orchestrates: script, voiceover, background music, video clips/images, and final assembly.
---
# Video Producer
Create complete videos with voiceover, music, and visuals.
**This is an orchestrator skill** that combines:
- Script/storyboard generation (Claude)
- Voiceover synthesis (Gemini TTS)
- Background music (Lyria)
- Video clip generation (Veo 3.1) or image animation
- Final assembly (FFmpeg via media-utils)
## Workflow
### Step 1: Gather Requirements (REQUIRED)
⚠️ **DO NOT skip this step. DO NOT run init_project.py until you have ALL answers.**
**Use interactive questioning** — ask ONE question at a time, wait for the response, then ask the next. This creates a collaborative spec-driven process.
#### Question Flow
⚠️ **Use the `AskUserQuestion` tool for each question below.** Do not just print questions in your response — use the tool to create interactive prompts with the options shown.
**Q1: Subject**
> "I'll create that video! First — **what's it about?**
>
> *(e.g., product launch, brand story, tutorial, explainer — or describe your own)*"
*Wait for response.*
**Q2: Duration**
> "How long should the video be?
>
> - 15 seconds *(quick hook)*
> - 30 seconds *(standard ad)*
> - 60 seconds *(explainer)*
> - 2+ minutes *(detailed)*
> - Or specify your own duration"
*Wait for response.*
**Q3: Style**
> "What visual style?
>
> - Premium/luxury
> - Fun/playful
> - Corporate/professional
> - Dramatic/cinematic
> - Minimal/clean
> - Or describe your own style"
*Wait for response.*
**Q4: Assets**
> "Do you have existing images or video clips to use?
>
> - No, generate everything
> - Yes, I have images *(provide paths)*
> - Yes, I have video clips *(provide paths)*"
*Wait for response.*
**Q5: Audio Strategy**
> "How should we handle audio?
>
> - **Custom** — I generate voiceover + background music
> - **Veo native** — Use Veo's built-in dialogue/SFX/ambient
> - **Silent** — No audio, add later"
*Wait for response.*
**Q6: Voice** *(if custom audio)*
> "What voice tone for the voiceover?
>
> - Professional
> - Friendly/warm
> - Energetic
> - Calm/soothing
> - Dramatic
> - Or describe your own tone"
*Wait for response.*
**Q7: Music** *(if custom audio)*
> "What music vibe?
>
> - Modern electronic
> - Cinematic/epic
> - Upbeat pop
> - Ambient/chill
> - Corporate
> - Or describe your own style"
*Wait for response.*
**Q8: Format**
> "What **aspect ratio**?
>
> - 16:9 (YouTube, web)
> - 9:16 (TikTok, Reels, Shorts)
> - 1:1 (Instagram feed)"
*Wait for response.*
**Q9: Resolution**
> "What **resolution**?
>
> - 720p (faster generation)
> - 1080p (standard HD)"
*Wait for response.*
**Q10: Model**
> "Which **Veo model**?
>
> - `veo-3.1` — Latest, highest quality (default)
> - `veo-3.1-fast` — Faster generation, slightly lower quality
> - `veo-3` — Previous generation
> - `veo-3-fast` — Previous gen, faster"
*Wait for response.*
#### Quick Reference
| Question | Determines |
|----------|------------|
| Subject | Scene content and prompts |
| Duration | Scene count (Veo clips must be 4, 6, or 8 seconds) |
| Style | Visual prompts and music selection |
| Assets | Generate vs use existing |
| Audio | custom, veo_audio, or silent |
| Voice | TTS voice selection |
| Music | Lyria prompt |
| Format | Aspect ratio for Veo |
| Resolution | 720p or 1080p output quality |
| Model | veo-3.1, veo-3.1-fast, veo-3, veo-3-fast |
---
### Step 2: Initialize Project
Once you have the user's answers, initialize the project with their preferences:
```bash
python3 ${CLAUDE_PLUGIN_ROOT}/skills/video-producer-agent/scripts/init_project.py \
--name "Product Launch Vide在 GitHub 阅读完整来源 (打开外部页面)