Google Gemini adds automated transcription and filler word removal
Google is refining its voice-processing capabilities with the introduction of Gemini 3.5 Transcribe, a model designed to strip out verbal stumbles like "ums" and "ahs" in real time. The update, rolling out to macOS and Android users, aims to replace manual editing with intelligent, context-aware speech recognition.
The new model moves beyond basic speech-to-text by allowing users to upload custom vocabularies, ensuring specialized jargon and unique terminology remain intact during processing. Beyond cleaning up filler words, the system supports over 85 languages and can distinguish between three distinct speakers in a single recording, providing word-level timestamps for precise reference.
Alongside the transcription tool, Google is deploying Gemini 3.5 Live and 3.5 Live Experimental. These versions focus on voice chat responsiveness, allowing the AI to handle mid-sentence interruptions more gracefully than its predecessor, Chirp 3. While the experimental iteration narrates its reasoning process in real time, the core suite aims to stabilize performance across complex tasks. These features are currently available for macOS app users in English, with the Rambler dictation tool on Android following in select markets.
Comments (0)
No comments yet. Be the first!