Installation¶
What you need¶
- Python 3.12 or newer (download)
-
uv — fast Python package manager. Install with:
-
An LLM API key — any of:
- Google Gemini (free tier available — recommended for getting started)
- Any OpenAI-compatible endpoint (OpenAI, Anthropic via proxy, local Ollama / LM Studio, etc.)
Optional, but unlocks more features¶
| Tool | Used by | When you need it |
|---|---|---|
| FFmpeg (download) | Subtitle, Voice, Dubbing, Live | Any audio/video workflow |
| LibreOffice (download) | Office formats on Linux/macOS | Translating legacy .doc / .xls / .ppt, or any Office file when MS Office isn't installed |
| Tesseract (install guide) | OCR engine (default) | Extract Text page, scanned-PDF translation, embedded-image translation |
| MS Office + pywin32 | Office on Windows | Highest fidelity Office translation on Windows |
You can install AI Translate without any of these — features that need them will tell you so before they fail.
Set it up¶
That installs everything needed to run the desktop app, the CLI, and the MCP server.
Run it¶
Add your API key¶
The first time you open the desktop app:
- Click Settings in the sidebar
- Open the LLM tab
- Paste your Google Gemini API key (or configure a custom OpenAI-compatible provider). Enterprise users can flip Gemini to Vertex AI mode instead — point it at a GCP project and region, optionally supply a service-account JSON path; see LLM Providers for the details.
- Pick a default model — any current Flash variant (e.g.
gemini-2.5-flash) is a solid free starting point. Pro variants give better quality at higher cost. - Close Settings — you're done
Keys are stored in your OS keychain (macOS Keychain, Windows Credential Manager, GNOME / KDE Secret Service on Linux), not in plain text on disk.
Headless / server install
If you can't run the desktop app to set up keys, see LLM Providers for the keychain CLI commands.