Gemini-3.5-Transcribe
Google is putting its newest transcription model into both developer APIs and everyday voice input surfaces.
Gemini 3.5 Transcribe is in public preview through the Gemini API, with separate live streaming and pre-recorded audio options. Google says it handles filler-word cleanup, self-corrections, custom vocabulary, more than 85 languages, and speaker attribution for recorded audio. The company cites Artificial Analysis results of 4.0% WER for streaming and 2.6% for non-streaming, plus a 70% improvement in time to final transcription over Chirp 3. It is also already tied into the Gemini macOS app and Rambler on Android, with Chrome support planned. HN · Frontpage AI's note
Gemini 3.5 Transcribe is in public preview through the Gemini API, with separate live streaming and pre-recorded audio options. Google says it handles filler-word cleanup, self-corrections, custom vocabulary, more than 85 languages, and speaker attribution for recorded audio. The company cites Artificial Analysis results of 4.0% WER for streaming and 2.6% for non-streaming, plus a 70% improvement in time to final transcription over Chirp 3. It is also already tied into the Gemini macOS app and Rambler on Android, with Chrome support planned. HN · Frontpage AI's note
score 6
Discussions
- hn · 253 points · 79 comments
- hn · 259 points · 82 comments
- hn · 267 points · 84 comments
- hn · 279 points · 85 comments
- hn · 283 points · 89 comments
- hn · 292 points · 92 comments
- hn · 304 points · 98 comments
- hn · 308 points · 101 comments
- hn · 314 points · 102 comments
- hn · 319 points · 105 comments
- hn · 321 points · 108 comments
- hn · 323 points · 109 comments
- hn · 327 points · 117 comments
- hn · 332 points · 119 comments