English | 简体中文

TalkingCut is a text-driven editing tool designed specifically for spoken video content.

Fast: Quickly remove unwanted footage through text editing.
Precise: Achieve millisecond-level refinement with waveform visualization.
Efficient: Dramatically optimize rough-cut workflows, letting creators focus on content.

• ✂️ Text-Driven Editing: Delete a word, phrase, or sentence in the editor, and the video is instantly "cut" to match.
• 🤖 Local AI Engine: Powered by WhisperX for high-precision, word-level timestamp alignment. Everything runs locally on your machine—no cloud, no privacy concerns.
• 📊 Waveform Refinement: Visualize audio waveforms for millisecond-level precision. The waveform syncs with the text editor in real-time—click on text to jump to the corresponding waveform position for fine-grained editing control.
• ⚡ Efficient FFmpeg Rendering: Utilizes FFmpeg's stream-copy and complex-filter capabilities for fast exports.
• 🍎 Silicon Optimized: Specifically architected to leverage Apple Silicon (MPS) for high-speed AI inference on MacBook hardware.

• Frontend: Next.js 15 (React) + Tailwind CSS + Lucide Icons.
• Desktop Wrapper: Electron (Node.js Main Process).
• AI Core: Python 3.10+ with WhisperX, Faster-Whisper, and Silero VAD.
• Video Engine: FFmpeg (integrated via Node.js child processes).

1. Import: Drop a video file into the app.
2. Transcribe: The local AI engine generates a word-level transcript with exact timestamps.
3. Edit: Review the text. Highlight "uhm"s or mistakes and hit Delete.
4. Preview: Play back the video in the app; it automatically skips the "deleted" segments.
5. Export: FFmpeg concatenates the kept segments into a new, polished video file.

This project is currently in active development. We are focusing on refining the AI alignment accuracy and the "breathing room" logic.

Check out the full development plan here: 👉 View the Development TODO List

Follow these steps to set up the development environment on your machine.

• Node.js: v18 or later
• pnpm: npm install -g pnpm
• Python: 3.11 or later
• FFmpeg: Must be available in your system PATH.
  • macOS: brew install ffmpeg
  • Windows: choco install ffmpeg

1. Clone the repository:

   git clone https://github.com/one-more/talkingcut.git
   cd talkingcut

2. Install Node.js dependencies:

   pnpm install

3. Set up Python Virtual Environment: This step installs the AI engine (WhisperX, PyTorch, etc.). This may take several minutes as it downloads large machine learning libraries.

   pnpm run python:setup

Start the Electron app in development mode:

pnpm run electron:dev

When the app opens:

1. Select a model size from the sidebar. Note: Larger models provide higher recognition accuracy but require more VRAM/RAM and are slower to process. If your device has sufficient resources (e.g., 16GB+ RAM), choosing a larger model is recommended. Medium is generally the best balance between speed and accuracy for most users.
2. Click the Download button or simply click "Open Video File"—the app will automatically download the required model weights to your local storage.
3. Once the status turns green (Cached), you're ready to transcribe!

Created with ❤️ by Jack.