Boneyard Tools

AI Translator

Translate text between 100 languages entirely in your browser. Unlike cloud translators, nothing you type is ever uploaded: the M2M100 AI model downloads once on first use, then runs locally and works offline. Pick a source and target language, paste your text, and translate. Your words stay on your device.

How to translate text privately

  1. Pick the source language your text is written in.
  2. Pick the target language you want it translated into (use the swap button to flip them).
  3. Paste your text and click Translate, then copy the result.

Examples

English to Spanish

Hello, how are you?
Hola, ¿cómo estás?

English to French

Good morning
Bonjour

Frequently asked questions

Is my text uploaded or sent to a server?

No. Nothing is uploaded. Unlike cloud translators such as Google Translate or DeepL, this tool runs the AI model entirely in your browser, so your text is translated on your own device and never leaves it. Only the model itself is downloaded, once, then cached.

Does it work offline?

Yes, after the first run. The model downloads the first time you use the tool and is then cached by your browser. From then on, translation works completely offline with no network connection needed.

Which AI model powers the translation?

It uses M2M100, a multilingual translation model that can translate directly between 100 languages without routing through English. The tool offers around 30 of the most common languages, all running locally in your browser.

How does the quality compare to Google Translate or DeepL?

M2M100 is genuinely good across many language pairs and is great for understanding text, drafts, and everyday translation. It is not perfect: for high-stakes or nuanced text, large cloud services can still be more polished. The trade-off here is complete privacy, since nothing is uploaded.

Why is the first run slow, and how is long text handled?

The first run downloads the model, which is a sizable one-time download (cached afterward for instant reuse). Long text is translated paragraph by paragraph to stay within the model's limit, so very long inputs simply take a little longer.

Related tools