Google adds Gemini 3.5 Transcribe for cleaner audio transcripts

In short: Google released new Gemini 3.5 Audio models, including a new tool called Gemini 3.5 Transcribe that turns audio into text and can automatically clean it up.

What happened

Google has updated its Gemini Audio lineup with new Gemini 3.5 models. The update includes Gemini 3.5 Live, Gemini 3.5 Live Experimental, and a new model called Gemini 3.5 Transcribe.

Google says these models are meant to improve how its voice features work in real life, like when there is background noise or when someone pauses or gets interrupted mid sentence. Think of it like a better listener that can still follow the conversation in a crowded room.

Gemini 3.5 Transcribe focuses on turning recorded speech into written text. Google says it can handle more than 85 languages and can better recognize specialized terms. It can also remove filler words like “um” and “uh,” add formatting, and even let people “edit naturally with just your voice.”

For people working with technical or brand specific words, Google says you can give the system a custom vocabulary so it spells terms the way you want. It can also label who is speaking for up to three different speakers and add timestamps for each word.

Google also noted that Gemini 3.5 Transcribe is a step up from its earlier transcription model, Chirp 3. This launch comes while Google is still expected to release Gemini 3.5 Pro, which it previously said would roll out in June.

Why it matters

Audio transcripts are used for meetings, interviews, podcasts, and customer support calls. A tool that better handles accents, multiple languages, interruptions, and common filler words can save time, because people spend less effort fixing the text afterward.

Source: The Verge AI