interflow-video-cut 是做什么的？
Automatically transform a local talking-head video into a card-style finished video: extract audio → transcribe with ElevenLabs → AI writes HTML cards per segment → render MP4. 10 visual styles × 4 layouts, transcription uses ElevenLabs only, never local models.
How It Works
- Extract audio from the input video file.
- Transcribe the audio using ElevenLabs' API (no local models involved).
- Segment the transcription by logical breaks or sentences.
- Generate HTML cards for each segment, styled based on the chosen visual style and layout.
- Render all cards into an MP4 video with transitions.
Features
- 10 Visual Styles: Choose from a variety of aesthetic themes (e.g., modern, minimalist, bold, vibrant).
- 4 Layouts: Options for card arrangement (e.g., single column, grid, side-by-side).
- ElevenLabs-Only Transcription: No local models required, ensuring consistent quality and speed.
- AI-Driven Design: The agent writes custom HTML for each card based on the spoken content.
Supported Platforms
- Claude Code
- Cursor
- Codex
- Other AI agents supporting skill/plugin architectures
Requirements
- ElevenLabs API key (set as environment variable
ELEVENLABS_API_KEY) - Input video file (MP4, MOV, or other common format)
- Node.js 18+ or Python 3.9+ (depending on runtime)
Usage
interflow-video-cut --input my_video.mp4 --output result.mp4 --style modern --layout single-column
Example
{
"input": "speech.mp4",
"output": "cards_final.mp4",
"style": "vibrant",
"layout": "grid"
}
Limitations
- Requires active internet connection for ElevenLabs API calls.
- Transcription accuracy depends on audio clarity and ElevenLabs service.
- Large videos may take longer to process due to transcription and rendering steps.
Credits
Created by derek-zhuolin. Contributions and feedback welcome.