ToolsAugust 26, 2026via The Verge AI
Google’s new AI transcription edits out your ‘ums’ and ‘ahs’
Why it matters
Google is shipping new transcription capabilities in Gemini Audio that handle real-world audio challenges (background noise, speech interruptions, disfluencies) and multilingual jargon detection. Practitioners using voice AI or building transcription workflows now have a stronger alternative; the automatic cleanup of ums/ahs saves post-processing time.
Key signals
- Gemini 3.5 Transcribe is a new model addition (not yet widely available based on article framing)
- Supports 85+ languages with specialized jargon detection
- Automatic disfluency removal (ums, ahs, stutters)
- Handles background noise and interrupted speech
- Ships as part of Gemini Audio update alongside Gemini 3.5 Live and 3.5 Live Experimental
- Gemini 3.5 Pro model still pending (promised June, not yet shipped as of Aug 2026)
- Gemini 3.5 Transcribe is a new model addition (separate from Live and Live Experimental variants)
- Automatically detects and filters filler words ('ums', 'ahs')
- Improved robustness to background noise and speech interruptions
- Part of Gemini Audio suite for voice-controlled AI
- Ships before Gemini 3.5 Pro release (promised June, delayed)
The hook
Google ships Gemini 3.5 Transcribe with automatic disfluency removal and 85-language support—live transcription gets a precision upgrade.
Google has updated Gemini Audio with some new Gemini 3.5 models, introducing new transcription capabilities that automatically detect specialized jargon and more than 85 languages. Gemini 3.5 Live, 3.5 Live Experimental, and 3.5 Transcribe are designed to provide better precision for Google's voice-…