I'M TranslAID
Local AI Setup

Run models on your own machine

Your text never leaves your computer. No API costs. Works offline. Pick a backend, download a model, start the server, point PTI at it.

Windows 10 / 11 · I'M TranslAID PTI already installed

Before you start
  • Choose either Ollama or LM Studio — you do not need both.
  • No API key required for local backends. If you have one saved from a cloud provider, leave it in place — it is ignored when a local backend is selected.
  • Local models vary in quality. Test on a sample file before committing to production use.
  • The local server must be running before you open PTI.
1 Install

A simple command-line tool for running local LLMs. Recommended if you are comfortable with a terminal.

Download Ollama ↗
2 Download a model

Open a terminal (Win + R → type cmd → Enter) and run ollama pull <model-name>. Browse available models at ollama.com/search.

Your hardware Recommended model size
No GPU / CPU only 3–4B parameters
8 GB GPU memory or more 7–12B parameters
Tag-heavy segments may challenge smaller models — if you see corrupted or missing tags in the output, try a larger model or switch to a cloud provider.
3 Start the server

Ollama runs automatically in the background after installation. Confirm it is active by opening a terminal and typing:

ollama list
4 Configure PTI

Open PTI and go to Settings (gear icon).

Click the Ollama preset button. The base URL is filled in automatically: http://localhost:11434/v1

Click Fetch models to load available models from your local server, then select the one you want to use.

Save settings and translate as normal.

If you have an API key saved from a cloud provider, leave it in place — it is ignored when a local backend is selected, and will still be there when you switch back.

1 Install

A desktop app with a model browser and GUI controls. Recommended if you prefer a graphical interface.

Download LM Studio ↗
2 Download a model

Open the Discover tab and search for a model.

Your hardware Recommended model size
No GPU / CPU only 3–4B parameters
8 GB GPU memory or more 7–12B parameters
Tag-heavy segments may challenge smaller models — if you see corrupted or missing tags in the output, try a larger model or switch to a cloud provider.
3 Start the server

Open LM Studio, load your model, and click Start Server in the Local Server tab.

The server must be running before you open PTI.

4 Configure PTI

Open PTI and go to Settings (gear icon).

Click the LM Studio preset button. The base URL is filled in automatically: http://localhost:1234/v1

Click Fetch models to load available models from your local server, then select the one you want to use.

Save settings and translate as normal.

If you have an API key saved from a cloud provider, leave it in place — it is ignored when a local backend is selected, and will still be there when you switch back.