AutoEditor

AI-powered video editing automation tool that automatically creates video timelines based on script matching using Whisper transcription and LLM analysis.

Features

• 🎬 Single Video Mode: Process a single video file with script alignment
• 🎥 Multicam Mode: Synchronize multiple video sources by audio waveform
• 🎤 Audio Transcription: Automatic transcription with word-level timestamps using Whisper
• 🤖 LLM Analysis: Intelligent script-to-transcription matching using OpenAI
• 📝 FCPXML Export: Generate Final Cut Pro XML timelines for DaVinci Resolve
• 🎨 Beautiful UI: Modern dark-themed web interface with native file pickers

Requirements

• Python 3.12 or higher
• uv package manager
• macOS (for native UI file pickers)
• OpenAI API key (for LLM analysis)

Installation

1. Install uv (if not already installed):

   curl -LsSf https://astral.sh/uv/install.sh | sh

2. Clone or navigate to the project directory:

   cd autoeditor

3. Install dependencies:

   uv sync

4. Create .env file in the project root:

   cp .env.example .env  # if you have an example file
   # or create manually:

   Add your OpenAI API key:

   OPENAI_API_KEY=your_api_key_here
   ELEVENLABS_API_KEY=...
   
   # Optional: Customize defaults
   SILENCE_THRESHOLD_DB=-40.0
   PRE_CUT_BUFFER=0.1
   POST_CUT_BUFFER=0.1

Usage

GUI Mode (Recommended)

Launch the beautiful web-based UI:

uv run autoeditor-ui

The application will open in a native window at http://127.0.0.1:8000 (or next available port).

Features:

• Select video files through native file picker
• Choose output directory
• Configure processing options
• Real-time progress tracking
• Live log output

CLI Mode

Single Video Mode

# Basic usage
uv run autoeditor video.mp4 --script script.txt

# With custom output directory
uv run autoeditor video.mp4 --script script.txt --output-dir output/my_project

# Export FCPXML timeline
uv run autoeditor video.mp4 --script script.txt --export

# Custom options
uv run autoeditor video.mp4 \
  --script script.txt \
  --bitrate 256k \
  --silence-threshold -35.0 \
  --pre-cut-buffer 0.2 \
  --post-cut-buffer 0.2 \
  --export

Multicam Mode

# With multiple videos
uv run autoeditor --videos cam1.mp4 cam2.mp4 cam3.mp4 --script script.txt

# With separate audio reference
uv run autoeditor \
  --videos cam1.mp4 cam2.mp4 \
  --audio reference_audio.wav \
  --script script.txt \
  --export

# Export multicam timeline
uv run autoeditor \
  --videos cam1.mp4 cam2.mp4 \
  --audio reference_audio.wav \
  --script script.txt \
  --export \
  --output-dir output/multicam_project

Command Line Options

General Options

• --script PATH: Path to script text file (required for LLM analysis)
• --output-dir PATH: Reuse existing output directory (skips transcription if JSON exists)
• --bitrate RATE: Audio bitrate (default: 192k)
• --start-time SECONDS: Start time for transcription processing
• --end-time SECONDS: End time for transcription processing

Export Options

• --export: Export FCPXML timeline from last_attempts.json
• --no-trim: Disable loudness-based end time trimming
• --silence-threshold DB: Silence threshold in dBFS (default: -40.0)
• --pre-cut-buffer SECONDS: Seconds to add before each cut (default: 0.1)
• --post-cut-buffer SECONDS: Seconds to add after each cut (default: 0.1)
• --original-audio: Use original audio file in timeline (default: use MP3)

Workflow

1. Transcription: Video/audio is converted to MP3 and transcribed with Whisper
2. Analysis: LLM matches transcription to script and identifies "last attempts" (best takes)
3. Timeline Generation: FCPXML timeline is created with cuts at identified moments
4. Export: Timeline can be imported into DaVinci Resolve or Final Cut Pro

Output Structure

output/
└── YYYYMMDD_HHMMSS/
    ├── source.mp3 (or audio.mp3 for multicam)
    ├── source_transcription.json (or audio_transcription.json)
    ├── last_attempts.json
    ├── llm_log.jsonl
    └── timeline.fcpxml (or timeline_multicam.fcpxml)

Development

Type Checking

# Run mypy
uv run mypy .

# Run pyright
uv run pyright

Project Structure

autoeditor/
├── src/
│   ├── main.py          # CLI entry point
│   ├── ui.py            # GUI entry point
│   ├── transcription.py  # Whisper transcription
│   ├── llm.py           # OpenAI LLM analysis
│   ├── sync.py          # Audio waveform synchronization
│   ├── fcpxml.py        # FCPXML generation
│   ├── video.py         # Video processing
│   └── files.py         # File management
├── output/              # Generated outputs
├── pyproject.toml       # Project configuration
└── README.md           # This file

Troubleshooting

UI doesn't open file pickers

• Make sure you're running in native mode (default)
• Check that pywebview is properly installed: uv sync

Transcription fails

• Verify your OpenAI API key is set in .env
• Check that video/audio file is accessible and valid

Multicam sync issues

• Ensure all video files have audio tracks
• Try using a separate high-quality audio reference file

License

[Add your license here]

Contributing

[Add contribution guidelines here]