The Built-in Translation API (also known as the Chrome Built-in AI Translator) is a client-side translation engine integrated directly into the Chrome browser (version 131+). Unlike cloud-based services such as Google Cloud Translation or Microsoft Translator, all translation processing happens on your device — no data is sent to external servers. This ensures complete privacy, works offline after model download, and eliminates latency from network round-trips. It's part of Chrome's broader Built-in AI initiative that brings machine learning capabilities directly into the browser.

1. Make sure you're using Chrome 131 or newer (check at chrome://version).
2. Navigate to chrome://flags/#translation-api in your address bar.
3. Set the "Translation API" flag to "Enabled".
4. You may also need to enable chrome://flags/#optimization-guide-on-device-model for the underlying AI model.
5. Click "Relaunch" to restart Chrome.
6. Return to this page — the status indicator should turn green once the API is ready.

The Built-in Translation API supports 40+ languages with English as the central pivot. Common pairs include: English ↔ Spanish, Japanese, Chinese (Simplified), French, German, Korean, Portuguese, Russian, Hindi, Arabic, Italian, Vietnamese, Thai, Turkish, Polish, Dutch, Indonesian, Ukrainian, Czech, Swedish, Danish, Finnish, Norwegian, Hungarian, Greek, Hebrew, Romanian, and many more. Some non-English pairs are also supported (e.g., Spanish ↔ Japanese, French ↔ German). Use the language selectors above to explore available combinations — the tool will indicate if a specific pair needs a model download or isn't yet available.

Yes, absolutely. Once the language model is downloaded, all translation processing runs locally on your device's CPU or GPU. Your text never leaves your computer. This makes it ideal for translating sensitive documents, confidential business communications, personal messages, or any content you wouldn't want to share with third-party cloud services. The model files are stored locally in Chrome's internal storage. You can verify this by disconnecting from the internet after downloading models — translation continues to work perfectly offline.

Feature	Built-in API	Google Translate
Privacy	✅ Fully on-device	⚠️ Cloud-based
Offline	✅ After model download	❌ Requires internet
Latency	✅ Near-instant	⚠️ Network-dependent
Languages	40+ pairs	130+ languages
Cost	✅ Free, unlimited	⚠️ API has quotas
Context awareness	Basic	Advanced (Neural MT)

Each language pair model is approximately 15–40 MB in size. The exact size depends on the language pair complexity. When you first use a new language combination, Chrome downloads the necessary model files automatically. These models are cached locally, so subsequent translations in the same language pair are instant. You can manage downloaded models through Chrome's internal storage settings. The download typically completes in under a minute on a standard broadband connection.

Currently, the Built-in Translation API is a Chrome-specific feature (version 131+). It is not available in Firefox, Safari, or older versions of Edge. However, Microsoft Edge (which is Chromium-based) may support it in future versions if they adopt the same underlying infrastructure. Firefox has its own separate machine learning initiative but does not yet offer a comparable built-in translation API. For cross-browser compatibility, consider using a cloud-based translation service as a fallback alongside the Built-in API.

Yes! The Translation API is exposed to web developers through the window.translation interface in Chrome 131+. You can call await translation.canTranslate() to check availability, await translation.createTranslator() to get a translator instance, and await translator.translate() to perform translations. The API is available in both Window and Worker contexts. It's designed for progressive enhancement — you can feature-detect it and fall back to cloud APIs when unavailable. This tool itself is open-source and demonstrates the full API integration pattern.