Analysis updated 2026-05-18
Translate documents or text across 200+ languages without your content leaving your machine.
Highlight foreign-language text on any webpage and see an instant translation overlay in your browser.
Run a private translation service on a home server or air-gapped machine using Docker.
| toto-sys28/freetranslate | ijzereen/branch-of-thought | arata-ae/purupurupngtuber | |
|---|---|---|---|
| Stars | 5 | 5 | 4 |
| Language | JavaScript | JavaScript | JavaScript |
| Setup difficulty | moderate | easy | moderate |
| Complexity | 3/5 | 2/5 | 3/5 |
| Audience | general | general | general |
Figures from each repo's GitHub metadata at analysis time.
Requires Docker or a Python 3.8+ environment and downloads ~5 GB of model files on first run, GPU acceleration needs NVIDIA Container Toolkit.
FreeTranslate is a translation application that runs entirely on your own computer. It uses open-source AI models to translate text across more than 200 languages without sending anything to an external server. Once the models are downloaded, the app works offline. No API keys or accounts are required. The application has two parts: a web interface you open in a browser at a local address, and an optional browser extension. The web interface lets you type or paste text, select source and target languages, and translate with a button click. It also supports file translation for DOCX, PDF, and plain text files. The browser extension adds a separate layer: highlight any text on a webpage and a translation card appears automatically, or click the extension icon to translate an entire page at once. Under the hood the app uses Facebook's NLLB-200 models, which were built for low-resource languages and cover a wide range of scripts and dialects. The inference runs through CTranslate2, a library that converts these models into a faster format. If you have an NVIDIA graphics card, the app detects it and uses the GPU for noticeably faster translations. If not, it falls back to the CPU, which is slower but still functional. Installation comes in two flavors. The Docker path is the simpler one: clone the repo, run docker-compose up, and the app is available at localhost:8000. The manual path installs Python dependencies in a virtual environment and starts a local server with one command. The browser extension is loaded into Chrome, Edge, Firefox, or other browsers in developer mode by pointing the browser at the extension folder. The default model uses 1.3 billion parameters and needs at least 8 GB of RAM. A smaller 600-million-parameter model is available for machines with less memory. Both models are downloaded on first run. The app stores them in a local directory and reuses them on subsequent starts. The README says storage needs around 5 GB minimum for the model files.
A fully local, offline-capable translation app supporting 200+ languages via open-source AI models, with a web interface and browser extension, no API keys needed.
Mainly JavaScript. The stack also includes Python, JavaScript, CTranslate2.
The README does not specify a license.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Verify against the repo before relying on details.