Google has launched Gemini 3.5 Transcribe, with new transcription features, support for special terminology and more than 85 languages. The emergence of the new model comes against the backdrop of the delayed release of the flagship Gemini 3.5 Pro.
Image source: blog.google
According to Google, Gemini 3.5 Transcribe “A significant improvement over our previous Chirp 3 speech transcription model”, especially in terms of supporting multiple languages and reducing the number of errors in recognized text. It allows users to “Edit text naturally using only your voice”, automatically formats the transcript and removes filler words.
Users can connect their own dictionaries to the model, so Gemini 3.5 Transcribe will take into account special terms, proper names, and non-standard spellings of words, eliminating the need to manually correct them. When processing recorded audio, the model is able to separate the lines of up to three speakers and place a timestamp on each word.
Google is also preparing Gemini 3.5 Live and 3.5 Live Experimental, which will expand the capabilities of the speech recognition and processing technology underlying Gemini voice chat. The company initially told reporters that both models were launching at the same time as the Gemini 3.5 Transcribe. However, after the information was released, Google clarified that it has only released Transcribe, and the release date of Gemini 3.5 Live and 3.5 Live Experimental has not yet been determined. Previously, the company said Gemini 3.5 Live would be better able to cope with mid-sentence interruptions, on-the-fly language recognition and visual material processing, and Live Experimental would be able to comment on the progress of complex tasks.
Gemini 3.5 Transcribe supports English and is now available for users of the Gemini app for macOS. In addition, the new Rambler voice input feature will soon be launched on the Android version of Gboard in some countries and supports multiple languages. The model is also available for public preview to developers through the Gemini API in Google AI Studio and Antigravity. Chrome should support Gemini 3.5 Transcribe in the near future.
Supports over 85 languages, including Russian; the model can also automatically detect languages and switch between multiple languages within a record.
If you find an error, select it with your mouse and press CTRL+ENTER.
