qiaomu-cut

Describe the video you want in one sentence; qiaomu-cut turns it into a source-aware, renderer-ready, verifiable video project.

Use Cases

• Movie clip mashups / English learning videos
• 60-second science or educational videos (e.g., excavator English for kids)
• Biographical shorts / historical documentaries
• 3Blue1Brown-style math or science explainers
• Product launch videos with web animation
• Vertical short videos for TikTok, Xiaohongshu, etc.
• Talking-head edits / podcast clips
• And more (see built-in workflows)

Core Capabilities

1. Converts a one-sentence brief into a QiaoCut IR that captures duration, aspect ratio, audience, style, shot plan, and material strategy.
2. Automatically searches, generates, or organizes media: 33tc (film clips), ClipSeek (free stock), Pexels/Pixabay, local files, ListenHub (image/video/TTS/music), AI image generation, etc.
3. Designs professional shot types: push/pull, match cuts, J/L cuts, subtitle tracking, title cards, motion graphics.
4. Supports multiple rendering engines: ffmpeg-full, HTML, Manim, and others.
5. Provides preview/standard/final render profiles so you can iterate quickly and then produce the final deliverable, with automatic technical validation at each level.
6. Delivers the final video, contact sheet, asset manifest (assets-manifest.json), license report (license-report.md), render quality report, and more.

How It Works

1. You provide a one-sentence video brief.
2. The system generates a QiaoCut IR and selects the most appropriate workflow.
3. Media is gathered or generated according to the IR, with sources and license status recorded.
4. A project timeline is assembled, supporting three-layer bilingual subtitles, branded intro/outro templates, and multi-track audio.
5. The video is rendered via ffmpeg etc. using the chosen profile, and technical validation is performed.
6. The video and associated files are produced; the project remains fully reproducible for later adjustments and re-renders.

What You Need

• A clear video brief (topic, duration, style, platform, etc.)
• The required runtime environment (see notes below)
• For 33tc: a locally installed and logged-in 33tc App plus an authorized 33tc CLI adapter
• For ListenHub generation: an account or API key

What You Get

• Final video file (.mp4), optionally a contact sheet
• QiaoCut IR file, asset manifest, license report, render quality report
• A complete, reproducible project that you can adjust and re-render at any time

Usage Examples

Check your environment:

node ~/.agents/skills/qiaomu-cut/scripts/qcut.js doctor --json

Create a 60-second excavator English-learning video:

# Generate a plan
node scripts/qcut.js plan "Make a 60-second excavator English learning video" --workflow stock-story --json

# Scaffold the project
node scripts/qcut.js scaffold ./excavator-video --brief "Make a 60-second excavator English learning video" --json

# (edit timeline.json) Quick preview
node scripts/qcut.js render ./excavator-video --profile preview --json

# Final render after content approval
node scripts/qcut.js render ./excavator-video --profile final --json

Search movie quotes with 33tc:

node scripts/qcut.js 33tc search "dig deeper" --limit 8 --json

Generate ASS bilingual subtitles:

node scripts/qcut.js ass captions.json --output subtitles.ass

Render with a contact sheet:

node scripts/qcut.js render ./my-project --profile final --json

Render Profiles

Built-in Workflows

Material Sources

• 33tc: Search for film/TV dialogue and clips. Requires a locally installed and logged-in 33tc App and an independently installed, authorized 33tc CLI.
• ClipSeek: Free stock media discovery. Downloads ultimately happen on the original site (Pexels, Pixabay, etc.) after license verification.
• Pexels / Pixabay: Accessed via ClipSeek results; confirm download and license on the original platform.
• Local files: User-provided media takes priority and is never overwritten.
• AI generation: Covers, illustrations, backgrounds, gap B-roll, SVG/web visuals, etc.
• ListenHub / MarsWave: Images, short video clips, TTS, voices, multi-character podcasts, music.
• Information sources: Facts about people, companies, or events must be cited in the project.

Important Notes

Runtime Environment

• Node.js 18+ (20+ for ListenHub features)
• macOS recommended with Homebrew
• ffmpeg-full (with libass, drawtext, overlay, loudnorm, H.264 support)
  • Install: ~/.agents/skills/qiaomu-cut/scripts/bootstrap_macos.sh --install
  • You can set QIAOMU_FFMPEG environment variable to a custom path
• Optional: 33tc App and CLI; ListenHub account/API key; GitHub CLI; AI image generation tool (provided by the current agent environment)

Security & Limitations

• Your original media is never deleted or overwritten.
• API keys are never exposed in commands, logs, task captures, manifests, or Git; credentials are sanitized in terminal output.
• Remote generation may incur credit costs and always requires explicit --yes. Uploading local media additionally requires --allow-upload.
• No claim of commercial usability is made unless the original site’s license is explicitly recorded.
• AI-generated content is marked as ai_generated and never presented as real footage or archival material.
• Film clip usage is assumed for learning, commentary, research, or authorized contexts; you are responsible for clearing rights for public distribution.
• Planning status is never described as a finished render.

Platform Support

Environment Variables

Troubleshooting