Google has launched Gemini 3.5 Transcribe, its most precise speech-to-text model yet, replacing Chirp 3. It strips filler words, fixes self-corrections and formats text live, with a 4 percent word error rate on streaming audio and 70 percent faster final transcription. The model covers 85 languages, tags up to three speakers, powers Gboard’s Rambler on Pixel 11, and heads to Chrome soon for voice typing.
