# Émile

> Émile is an AI video editor for Mac with an agent inside. Describe the video, or bring your script; the agent, powered by Claude, Codex or Gemini, researches, plans and draws every clip on a real multi-track timeline, with AI voice-over, captions and motion graphics. Made for editors and creators of short videos and YouTube. Coming soon.

Émile (also written "Emile") is in development. There is no public release or download yet; the first public version will be published at https://emilefilm.com. It runs on Macs with Apple silicon and is open source under the Apache License 2.0.

Key facts:

- Category: AI agent video editor for short videos and YouTube, on macOS (Apple silicon).
- Made for: video editors (a first cut from the agent, finished on a real timeline) and content creators (short-form video for TikTok, YouTube Shorts, Instagram Reels, Douyin, Xiaohongshu, Bilibili, and long-form YouTube); also explainer channels and brands (animated charts, motion graphics, brand kits).
- Creator features: idea or script to video, AI voice-over in 13 languages on the Mac, voice cloning (ElevenLabs), captions in nine styles, music and sound effects, safe areas for TikTok, Shorts, Reels and YouTube, vertical 9:16 or wide 16:9.
- The agent: researches, plans and draws every clip in code (type, charts, diagrams and motion) onto a real multi-track timeline (picture, sound, voice-over), then keeps editing by chat, by drawing on the frame, or by hand. Every step can be undone.
- AI providers: Claude, Codex and Gemini only. Through their command-line tools (Claude Code, Codex CLI, Gemini CLI) on an existing subscription, or through an API key (Anthropic API, OpenAI API, Gemini API).
- Toolkits today: ElevenLabs (voice cloning, sound effects), GPT Image and Nano Banana (images, own key).
- On-device models (run on the Mac, no key): Whisper Turbo (transcription, captions), FLUX.1 schnell (images), RIFE (frame interpolation), BiRefNet (background removal), ACE-Step (music). Portrait cut-outs for photos and video use macOS's built-in person segmentation, with no model to download.
- In development: an AI rough cut from your footage. Whisper Turbo transcribes on the Mac with word-level timing; the agent splits chapters, picks highlights and lays out a first cut, and finds clips by what was said ("the part where he talks about price"). Auto-reframe from 16:9 to 9:16 keeps people in frame using macOS Vision face and body tracking. Best for talking-head and explainer videos.
- Planned video models: Seedance 2.5 and Kling 3.0.
- Export: MP4, WebM or MOV; 1080p or 4K; vertical 9:16 or wide 16:9.
- Voice-over is synthesised on the Mac: CosyVoice 3 by default, VoxCPM2 for higher quality, Kokoro as the English-only fallback; ElevenLabs is optional.
- Privacy: at most one anonymous count per day, never projects, text, files or keys; can be turned off.
- Made by Oliver Zhang (https://oliverz.dev; X @oliverzdev, YouTube @oliverzdev, Instagram oliverzdev, GitHub oliverzdev). Contact: oliverz.dev@gmail.com. Named after Émile Reynaud, who projected the first animated films in Paris in 1892.

## Pages

- [Émile (English)](https://emilefilm.com/): the product, screenshots, the agent, FAQ, status
- [Émile（中文）](https://emilefilm.com/zh/): 同一页面的中文版本

## Docs

- [Full text for LLMs](https://emilefilm.com/llms-full.txt): every word on the site, English and Chinese
- [Source code and README](https://github.com/oliverzdev/Emile): build from source, repository layout, project format

## Optional

- [Privacy notice](https://oliverz.dev/products/emile/privacy): everything Émile stores and sends
- [Author](https://oliverz.dev)
- [Oliver Zhang on X](https://x.com/oliverzdev)
- [Oliver Zhang on YouTube](https://www.youtube.com/@oliverzdev)
- [Oliver Zhang on Instagram](https://www.instagram.com/oliverzdev)
- [Oliver Zhang on GitHub](https://github.com/oliverzdev)