Google’s new AI transcription edits out your ‘ums’ and ‘ahs’
The Verge Jess Weatherbed ● Covered by 6 sources
Google updated Gemini Audio with new 3.5 models for transcription. They handle jargon, 85+ languages, and messy speech while Pro is still missing.
Based on reporting by The Verge, Jess Weatherbed — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google has pushed Gemini Audio forward with a set of new Gemini 3.5 models aimed at transcription and voice control. The pitch is simple: better precision when people use specialized jargon, talk over noise, or get interrupted mid-sentence.
The lineup now includes Gemini 3.5 Live, Gemini 3.5 Live Experimental, and Gemini 3.5 Transcribe. Google says these models are built to make its voice-controlled AI features more reliable, which is the sort of upgrade that matters most when software is supposed to listen in real rooms instead of clean demo videos.
The newest piece is Gemini 3.5 Transcribe, a fresh addition to the Gemini family. Google says the system can automatically detect specialized jargon and handle more than 85 languages, which puts a lot of weight on the transcription side of the product rather than just the usual chatbot polish.
And there’s a bit of timing friction here. Google is still waiting to ship the Gemini 3.5 Pro model it said would roll out in June. So while one part of the family is getting broader and more practical, another promised piece is still on the way.
My take — AI-written commentary, not fact-checked reporting
Google keeps doing this neat little trick where the plumbing gets better while the headline model stays in the waiting room. That’s not nothing, because transcription is where AI actually has to earn its keep. But the company could stand to ship the promised model before moving on to the next shiny layer of the same stack.
Read more about this at: The Verge