
New: Local AI (Ollama / LM Studio) A long-requested feature is here! You can now translate with an AI model running on your own PC.
Free and unlimited: no API key, subscription or quota.
Private: your text never leaves your computer. (If free backup services are enabled and the local server does not respond, the text is sent to them; turn off the backup option on the same page if you don't want that.)
Any model you like: works with Ollama, LM Studio or any OpenAI-compatible server.
To use it: API Settings → Local AI → Ollama or LM Studio button → Verify → Load models.
Download models from inside the app (Ollama)
The settings show a list of recommended models with their download size, the graphics memory they need, and a short description.
Models already on your PC are marked ✓ Installed; the others can be downloaded with one click, with a progress bar. When the download finishes, the model is selected automatically.
Your graphics card (NVIDIA) is detected, and a warning is shown for models that need more memory than it has. Free disk space is checked before downloading.
Good to know
The first translation can take 10–30 seconds while the model is loaded into your graphics card; after that it is much faster.
Speed and quality depend on the model and your graphics card. Models of 7B and above translate well; very small models may be weak in some languages.
OpenAI model selection
The OpenAI model field is now a drop-down list. The models your API key can access (including the newer ones released after gpt-4o) are listed automatically and can be selected. You can still type a model name by hand.
Improvements and fixes
When the settings window is maximized, the content no longer stretches across the whole screen; it stays at a readable width in the center. Description texts are slightly larger.
Fixed the right edge of the API Settings page (with its buttons) going off-screen in languages with long texts such as Russian and Ukrainian.
