Google’s new AI transcription edits out your ‘ums’ and ‘ahs’

Google has updated Gemini Audio pinch immoderate caller Gemini 3.5 models, introducing caller transcription capabilities that automatically observe specialized motto and much than 85 languages. Gemini 3.5 Live, 3.5 Live Experimental, and 3.5 Transcribe are designed to supply amended precision for Google’s voice-controlled AI features, without struggling pinch inheritance sound aliases erstwhile your reside is interrupted.

Gemini 3.5 Transcribe is simply a wholly caller summation to the Gemini family, and its preamble comes arsenic we’re still waiting for Google to merchandise the Gemini 3.5 Pro exemplary that it promised to rotation retired successful June. Google says that 3.5 Transcribe “represents a awesome advancement from our erstwhile transcription model, Chirp 3,” particularly regarding multilingual capacity and wording correction rates.

The transcription exemplary allows users to “edit people pinch conscionable your voice,” according to Google, and tin automatically format matter and region filler words for illustration “um” and “uh.” Users tin supply a customized vocabulary to the model, allowing 3.5 Transcribe to automatically accommodate transcription to unsocial pronunciation requirements and specialized motto to forestall those words from being edited manually. It tin besides property reside for up to 3 speakers successful pre-recorded audio, alongside providing word-level timestamps.

Transcribe is launching alongside 3.5 Live and 3.5 Live Experimental, which build connected the existing reside nickname tech that powers Gemini’s sound chat mode. Gemini 3.5 Live is amended astatine handling mid-sentence interruptions, connection recognition, and unrecorded ocular processing, while Gemini 3.5 Live Experimental goes further by narrating its advancement measurement by measurement successful existent clip while it tackles reasoning connected much analyzable tasks.

These Gemini Audio updates are rolling retired starting coming successful English for each macOS Gemini app users, and the Rambler dictation characteristic connected Android successful prime countries and languages. It’s besides disposable for developers successful nationalist preview successful the Gemini API via AI Studio and Antigravity. Google says that Chrome support is coming soon.