Gemini 3.5 Transcribe is Google's most precise speech-to-text yet
Google just rolled out Gemini 3.5 Transcribe, its most precise speech-to-text model yet.
This new model handles background noise, tricky tech terms, and filler words way better than before, and it even learns your personal vocab and speaking style for smoother edits.
It's already powering Gboard Rambler and will soon let you use your voice to type directly in Chrome.
Gemini 3.5 Transcribe supports 85+ languages
Gemini 3.5 Transcribe transcribes audio in more than 85 languages, catching different accents and up to three speakers at once (with timestamps).
It's also super quick, processing audio 70% faster than the last version, and nails accuracy with a low word error rate (4.0% streaming, 2.6% non-streaming).
Right now you can try it on Mac or through Google Antigravity; more platforms like Chrome are coming soon.