Google Translate AI

Powered by Google's transformer-based GNMT and successor models, translations are produced sentence-by-sentence rather than word-by-word, preserving grammar and intent across an unmatched language footprint that includes many low-resource and recently added languages.

Point your phone camera at signs, menus, or documents and Google Translate uses OCR plus on-device ML to render translated text in place, mimicking the original layout. Works in real time for live scenes and supports importing photos for batch translation.

Acts as a two-way interpreter — both speakers talk in their native language and the app transcribes, translates, and reads each turn aloud, enabling face-to-face communication without typing. Particularly useful for travel, customer service, and informal cross-language conversations.

Downloadable language models in the mobile apps enable text and camera translation without an internet connection. Quality is reduced compared to online but functional, making it practical for travel where roaming data is expensive or unavailable.

Built into Chrome (full-page translation), Google Search, Gmail, Google Docs, Google Lens, and Android system tap-to-translate, so translations are one click or long-press away in most everyday applications without launching a separate app.

Developer-facing API with Basic and Advanced tiers, both priced at $20 per million characters. Advanced supports custom glossaries, document translation that preserves formatting (PDF, DOCX), batch translation, and AutoML Translation for training custom domain models, enabling integration into commercial products and localization pipelines.