Gemini-3.5-Transcribe

Gemini 3.5 Transcribe converts raw audio to formatted text with a 4.0% word-error rate in streaming mode and 2.6% in non-streaming mode, handling noise, filler words, self-corrections, and custom vocabularies. It provides two APIs—real-time streaming with sub-second latency and batch processing with speaker attribution, word-level timestamps, support for 85+ languages and up to three speakers—and includes function calling for tasks such as image generation. The model is integrated into the Gemini macOS app, Gboard’s Rambler feature, and Android voice capabilities.