Google’s Gemini 3.5 Transcribe adds multilingual, jargon-aware transcription

Google updates Gemini Audio with Gemini 3.5 Transcribe, a multilingual transcription model that can “edit naturally with just your voice,” automatically format text and remove filler words like um and uh, and adapt to specialized jargon via a user-provided vocabulary. It can attribute speech for up to three speakers and provide word-level timestamps. Google also teased 3.5 Live and 3.5 Live Experimental updates, but later said only 3.5 Transcribe is being announced today; rollout begins today in English on macOS Gemini apps and Rambler dictation on Android in select markets, with Chrome support coming soon and API access in public preview.