AutoClipper

AI-powered desktop app that turns a long video into short, ready-to-post clips — automatic transcription, AI segment selection, smart vertical cropping, styled burned-in subtitles, and privacy face/head blurring, all running locally.

Built with Tauri (Rust + React/TypeScript), open source under the MIT license.

Download

Prebuilt installers (macOS .dmg, Linux .deb, Windows .zip) are on the Releases page.

Features

• Transcription — local Whisper transcription (faster-whisper, with mlx-whisper GPU/Neural Engine acceleration on Apple Silicon), multi-language source detection, speed presets (fast/balanced/accurate/best)
• Subtitle chunking — group transcript into subtitle lines by word count (auto / 1 / 2 / 3 words), including karaoke-style one-word captions
• Translation — translate the transcript to another language, with original-only / translated-only / bilingual subtitle modes
• AI segment selection — a local LLM (Ollama or a bundled GGUF model) classifies the video into sections and picks the most important segments, or select segments manually
• Draggable subtitle timeline — trim each subtitle's start/end time by dragging directly on a video-synced timeline, like CapCut/DaVinci Resolve
• Smart crop — automatically follow the speaker's face when converting to vertical (9:16), with smooth or aggressive tracking, plus fixed aspect ratios (16:9, 1:1, 4:5)
• Burned-in subtitles — customizable color, outline, background box, position, font, size, and ALL CAPS
• Face/head censoring — pixelate or overlay-image privacy blur, detecting faces or full heads (MediaPipe), for anyone who needs to stay off camera
• Caption generation — AI-generated short/long captions and hashtags per clip
• YouTube import — download a source video directly by URL (yt-dlp[default] + Deno as the JS-challenge runtime)
• Bilingual UI — Indonesian and English

Tech Stack

• App shell: Tauri 2.x (Rust)
• Frontend: React + TypeScript, Vite
• Processing: Python (faster-whisper/mlx-whisper, OpenCV, MediaPipe, InsightFace), FFmpeg
• Local AI: Ollama or a bundled GGUF model for segment classification and captions

Prerequisites

• Rust + Cargo
• Node.js 18+ and npm
• Tauri CLI prerequisites for your OS
• Python 3 (the app can set up its own bundled Python environment — see below)
• FFmpeg + FFprobe (bundled automatically on first run if not found on your system)
• Optional: Ollama for local LLM-powered segment selection and captions
• Optional: Deno for YouTube import — yt-dlp needs a JS runtime to solve YouTube's signature challenge; without it, downloads fail with HTTP 403

Development

npm install
npm run tauri dev

The app checks its Python/FFmpeg/model dependencies on first launch and offers to install what's missing (see scripts/setup_bundle.sh / scripts/setup_bundle.ps1).

Run the frontend test suite:

npm test

Building a Release

# macOS / Linux
./scripts/build_release.sh

# Windows
./scripts/build_release.ps1

This bundles Python, FFmpeg, and the required models into a self-contained app under src-tauri/target/release/bundle/.

Project Structure

src/              React/TypeScript frontend
src-tauri/        Rust backend (Tauri commands, process orchestration)
scripts/          Python processing scripts (transcribe, censor, crop, burn subtitles)
docs/superpowers/ Design specs and implementation plans for past features

License

MIT — see LICENSE.