ClipCutter

Audio-based video highlight extractor. Point it at a folder of gaming recordings and it automatically finds the good moments — volume spikes, laughter, shouting, sudden noises — and pulls them out as clips for you to review and export.

How it works

1. Process — analyzes audio to detect highlight events and extracts clips
2. Review — watch each clip, trim it, name it, keep or discard
3. Export — encode, build compilations, upload to YouTube

Features

Process tab

• Scans a folder and lists videos with size, age, and processing status
• Adjustable sensitivity (0.1–5.0) and context window around each highlight
• Detects: volume spikes, laughter, shouting, sudden noises
• Live processing log

Review tab

• Video player with interactive waveform
• Colored overlays show which detection type triggered each highlight region
• Trim controls: set in/out points per segment, add multiple segments
• Trim quality: Fast (stream copy), Precise (CRF 16), Ultra (lossless)
• Optional custom name per clip for use during export
• Keyboard shortcuts: K keep, D discard, S skip, I/O set in/out, Space play/pause

Export tab

• Disk usage summary across kept, encoded, and compilation folders
• Encode presets: Original (copy), High (H.264 CRF 18), Low (H.264 CRF 26), GIF (with optional slowdown)
• Build compilations with hard cut or crossfade transitions, drag to reorder
• Delete encoded versions individually to reclaim space
• Source video cleanup — delete originals once review is complete
• YouTube upload via OAuth2 with title, description, tags, category, and playlist

Requirements

• Python 3.10+
• FFmpeg (must be on PATH)

Installation

git clone https://github.com/curtyo18/ClipCutter.git
cd ClipCutter
pip install -e .

Usage

Web UI (recommended)

python -m clipcutter ui

Opens a browser at http://localhost:8000 with the Process, Review, and Export tabs.

On Windows you can also double-click ClipCutter.bat.

CLI

# Process a folder of videos
python -m clipcutter process H:\OBSReplays -o ./output

# Adjust sensitivity (higher = more clips)
python -m clipcutter process H:\OBSReplays -s 1.5

# Dry run — detect highlights without extracting clips
python -m clipcutter process H:\OBSReplays --dry-run

# Terminal-based review (no browser needed)
python -m clipcutter review -o ./output

Output structure

output/
  clips/pending/<video>/      # Awaiting review
  clips/kept/<video>/         # Approved clips
  clips/encoded/<video>/      # Re-encoded clips
  clips/compilations/         # Built compilations
  metadata/<video>_clips.json # Per-clip metadata

Stack

Python, FastAPI, librosa, numpy/scipy, FFmpeg, TypeScript

Running tests

pip install pytest httpx playwright
python -m playwright install chromium

pytest tests/ -v                   # Full suite
pytest tests/ -v -k "not browser"  # API tests only (faster)