Clipify
A Claude Code skill that turns long videos into social-ready clips.
Point it at any video file and it will:
- Find clip-worthy segments — transcribes the video with Whisper and scans the transcript for punchlines, reversals, awkward pauses, and audio peaks to propose 3–5 candidates.
- Create a 9:16 clip — cuts your chosen moment, then reframes 16:9 → 9:16 with hard-cut pans that follow whoever is speaking (or split-screen if you’d rather see both faces).
- Add subtitles — burns opus-style word-by-word captions (big bold white, yellow active-word highlight).
No cloud APIs. Runs entirely on your machine. No OpenCV. ~20s of work for a 20s clip on Apple Silicon.
Why this exists
Most “auto-clip” tools are either expensive SaaS, slow, or produce slop. This skill is what I actually use to clip my long-form videos for LinkedIn and TikTok. Built for talking-head dialogue (interviews, podcasts, two-person setups).
Requirements
- macOS (uses VideoToolbox for hardware-accelerated decode — works on Linux/Windows if you remove
-hwaccel videotoolbox flags)
- Claude Code
ffmpeg with libx264 (brew install ffmpeg)whisper (pip install openai-whisper)- Python 3 with
numpy (pip install numpy)
Install
git clone https://github.com/louisedesadeleer/clipify.git ~/.claude/skills/clipify
That’s it. Restart Claude Code and /clipify is available as a slash command.
Usage
In Claude Code:
/clipify
Then paste a video file path when asked. The skill will:
- Transcribe → propose 3–5 funny candidate clips with timestamps and titles
- Ask which to cut
- Ask 9:16 / 16:9 / 1:1
- If 9:16 from 16:9 with two faces: ask pan vs split-screen
- Ask subtitle style (opus / karaoke / minimal — or paste a reference image to match)
- Render and open the result
Final clips land in <source-video-dir>/clipify_out/.
How the face-pan works
No face detection model. Camera is static within a single clip, so:
- Eyeball each face’s mouth+chin area as a rectangle on one sample frame.
- ffmpeg computes per-frame motion energy in each rectangle using frame differencing.
- Whichever rectangle has more motion at a given moment = that’s the speaker.
- Build a hard-cut x-coordinate expression from the speaker timeline.
- Crop a vertical strip from the source that follows whoever’s talking.
Total cost: a few seconds of ffmpeg per clip. Works surprisingly well.
Repo structure
clipify/
├── SKILL.md # the skill prompt Claude Code reads
├── scripts/
│ ├── analyze.py # speaker timeline from two ROI motion files
│ ├── build_pan.py # ffmpeg crop x-expression with hard cuts
│ ├── build_ass.py # opus/karaoke/minimal ASS captions from whisper JSON
│ └── audio_align.py # find offset of a sub-clip in a longer source
└── README.md
License
MIT — see LICENSE.
Built by Louise de Sadeleer, Growth at Tella.
评论