Intelligent transcription with Gemini 3.5 Transcribe
deepmind ·
Google has unveiled Gemini 3.5 Transcribe, a new speech-to-text model designed for intelligent real-time transcription. It converts raw audio into polished, formatted text while handling background noise, complex jargon, and speech disfluencies. The model is available through two APIs: the Live API for streaming with sub-second latency, and the Interactions API for pre-recorded audio with speaker attribution and timestamps. Key features include smart transcription cleanup, function calling, custom vocabulary support, automatic detection of over 85 languages, and multi-speaker identification. According to the article, it achieves word error rates of 4.0% for streaming and 2.6% for non-streaming, improving on the previous Chirp 3 model. Early access is offered to developers via Google AI Studio, with integration planned for Gboard, Antigravity, the Gemini app, and Chrome.
deepmind