Google Launches Gemini 3.5 Transcribe Speech-to-Text Model

Google today launched the Gemini 3.5 Transcribe speech-to-text model. This model automatically removes filler words and self-corrections from spoken audio, producing smoother transcripts. It has already been integrated into the "Rambler" feature of the Gboard keyboard on Pixel 11 devices, with plans to expand to more Google products and services in the future. Compared to the previous Chirp 3 engine, Gemini 3.5 Transcribe offers approximately 70% faster transcription speeds and reduces real-time word error rates to 5.5%. The model supports 85 languages, can process pre-recorded audio with up to three speakers, and allows users to upload custom vocabulary lists to improve recognition accuracy for specialized terminology. During transcription, the model attempts to understand user intent and intelligently adjusts the text, which may alter the original phrasing.