AI Translate & Dub
Transcribe video with Whisper, translate subtitles via Ollama or DeepSeek, and export SRT, MKV, or hard-burned MP4 — all from a local Mac app. No account required. No subscription.
macOS · Apple Silicon (arm64) · Python 3 + ffmpeg required
What it does
Whisper transcription
Transcribe speech from any video using OpenAI Whisper. Choose from tiny, base, small, medium, or large models. Models are cached locally — no repeat downloads.
Two translation backends
Use a local Ollama model (100% private, no API key) or connect to DeepSeek for high-quality cloud translation. Switch anytime.
Global translation context
Optionally generate a topic, terminology, and style prompt from your video before translating. Improves consistency for technical, educational, or niche content.
Parallel translation
Translate multiple subtitle segments simultaneously. Adjust worker count to balance speed against API rate limits.
Three output formats
Export a standalone .srt file, an .mkv with a soft subtitle track (switchable in any player), or a hard-burned .mp4 ready for upload.
Runs on your Mac
Electron shell with a local Gradio UI. No cloud upload unless you use DeepSeek. Your videos never leave your machine by default.
How it works
- 01
Install the app
Download the DMG, drag AI Translate & Dub to Applications, and launch. The app starts a local Gradio server — your browser opens http://127.0.0.1:7860 automatically.
- 02
Upload your video
Drop any MP4 or MOV into the Gradio UI. Audio is extracted at 16 kHz mono.
- 03
Pick a Whisper model
Select tiny → large depending on your accuracy and speed needs. Model downloads once, then caches locally.
- 04
Choose a translation backend
Connect to a running Ollama instance or enter your DeepSeek API key.
- 05
Run and export
Transcription, translation, and muxing all happen in one click. Pick SRT, MKV, or MP4 output.
Output formats
| Format | Description | Best for |
|---|---|---|
.srt | Plain subtitle file with timestamps | Re-use in any editor or player |
.mkv | Original video + soft subtitle track | Local playback, subtitle toggle in VLC/IINA |
.mp4 | Hard-burned subtitles baked into video | Upload-ready for YouTube, social media |
Requirements
System
- macOS (Apple Silicon arm64 recommended)
- Python 3 installed on your machine
- ffmpeg (
brew install ffmpeg)
For local translation (Ollama)
- Ollama running locally
- Any compatible model pulled, e.g.
qwen3:latest - No internet required after setup
For cloud translation (DeepSeek)
- DeepSeek API key
- Internet connection
- Pay-per-use (no subscription)
FAQ
- Does it upload my video to the cloud?
- No. Video processing (audio extraction, Whisper transcription) all happens locally. Only the subtitle text is sent to DeepSeek if you choose that backend. Ollama is 100% local.
- Which Whisper model should I use?
baseis fast and accurate enough for most content. Usemediumorlargefor technical or heavily accented speech. Models are cached after the first download.- Can I translate into any language?
- Translation quality depends on your chosen backend model. Ollama models like
qwen3support Chinese, Japanese, Korean, and major European languages. DeepSeek covers 50+ languages. - Is Intel Mac supported?
- The current DMG is arm64 only. Intel support may be added in a future release. You can run from source on Intel now.
- Where are Whisper models stored?
- Models are cached to
~/Library/Application Support/AI翻译配音/whisper-models. If you have existing models in~/.cache/whisper, they are copied over automatically on first launch. - Is it free?
- Yes. The app is free and open-source (MIT). DeepSeek API usage is billed by DeepSeek directly at their standard rates.
Download AI Translate & Dub
v1.0.0 · macOS arm64 · Free & open source (MIT)