GoKawiilGoogle rolls out Gemini 3.5 Transcribe, a faster AI speech-to-text model
Google has launched Gemini 3.5 Transcribe, a new voice-to-text AI model that automatically strips out filler words and stumbles to produce cleaner transcriptions. Already running Gboard's Rambler feature on the Pixel 11, the model is set to expand across Google's products. Google says it is roughly 70 percent faster than its predecessor, Chirp 3, and slightly more accurate, with a live-speech error rate of 5.5 percent versus 7.32 percent.
GoKawiil's interpretation of the reporting above, not reported fact.
Voice input has long been hampered by messy transcripts full of false starts and typos that are tedious to fix by hand, so meaningful accuracy and speed gains matter for everyday usability. But because the model actively rewrites what users say rather than transcribing verbatim, it raises questions about fidelity in contexts like legal, medical, or professional dictation where exact wording counts.
- Gemini 3.5 Transcribe already powers Gboard's Rambler feature on Pixel 11 and will expand further.
- It's about 70% faster than Chirp 3, with a live-speech error rate of 5.5% versus 7.32%.
- The model edits speech in real time, removing filler words and correcting course, across 85 languages.
Blue Yeti USB Microphone — If you're going to lean on AI speech-to-text like Gemini 3.5 Transcribe, clean audio input makes a huge difference in accuracy. The Blue Yeti is a plug-and-play USB condenser mic with multiple pickup patterns, so you can dictate at your desk or record interviews without wrestling with drivers.
See Blue Yeti USB Microphone on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: arstechnica.com, 2026-08-26
Published there as: “Google announces Gemini 3.5 Transcribe for AI-powered speech-to-text”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.