A cross-platform desktop application that translates files into 40+ languages — documents, PDFs, subtitles, images, audio, and live speech.
📖 Full documentation: cadic2603.github.io/ai-translate
- Documents — DOCX / XLSX / PPTX / ODT / PDF / EPUB / HTML / Markdown / subtitles / localization files, with full formatting, hyperlinks, comments, headers/footers, footnotes, and embedded images preserved.
- Multiple LLM providers — Gemini (Developer API or Vertex AI) plus any OpenAI-compatible endpoint (OpenAI, Azure, Anthropic, DeepSeek, OpenRouter, Groq, vLLM, Ollama, LM Studio, …) via a 4-variant payload auto-fallback.
- Speech — Subtitle generation (Whisper / Google Cloud / Soniox), text-to-speech (Edge / Google Cloud / Gemini / Piper offline / ElevenLabs), end-to-end video dubbing, and real-time live translation with a floating subtitle overlay.
- Headless —
aitCLI for terminal use andait-mcpMCP server for AI agents (Claude Desktop, Claude Code, …). - Resumable — Checkpoint system so paused or crashed tasks resume from the last completed stage; atomic output means partial files never appear in your output folder.
Requires Python 3.12+ and uv.
git clone https://github.com/cadic2603/ai-translate.git
cd ai-translate
uv sync
uv run ai-translateFor CLI and MCP usage:
uv run ait report.docx --target French
uv run ait-mcp # stdio transport for Claude Desktop / Claude CodeSee the Getting Started guide for installation details, provider setup, and per-feature walkthroughs.
Python 3.12 · PySide6 · SQLite · uv · Ruff · pytest
Conventions for human and AI contributors live in AGENTS.md. Bug reports and feature requests are welcome at Issues.