Google updated its Gemini Audio transcription capabilities on Wednesday with the release of Gemini 3.5 Transcribe. The new model brings structured formatting, jargon recognition, and support for more than 85 languages directly to voice input systems. According to reporting from The Verge, the release aims to improve transcription quality across multiple platforms.
New Features in Gemini Audio Transcription
The updated model automatically cleans recorded speech by removing filler words such as “um” and “uh.” Furthermore, the software formats raw speech into structured paragraphs without requiring manual punctuation commands. Google stated that 3.5 Transcribe represents a major advancement over its previous Chirp 3 model, particularly in reducing word error rates.
“Gemini 3.5 Transcribe represents a major advancement from our previous transcription model, Chirp 3.”
Google
In addition, the system includes speaker attribution for up to three distinct voices in pre-recorded audio files. The tool also generates word-level timestamps to assist with precise editing in apps and media workflows.
Custom Vocabulary and Multilingual Capabilities
Users can submit custom vocabulary lists to the model before processing audio recordings. Consequently, the software recognizes specialized technical terms, product names, and unique spellings without misinterpreting them. This functionality prevents users from spending extra time making manual corrections after converting speech to text through artificial intelligence systems.
Multilingual performance is another central upgrade in this release. The model processes speech in more than 85 languages, allowing international teams to transcribe mixed conversations accurately.
Availability Across Platforms and Developer Access
Google has started rolling out 3.5 Transcribe in English for users of the macOS desktop application on computers. Meanwhile, mobile users can access the tool through the Rambler dictation feature on Android mobile devices in select regions. Developers can test the model in public preview through the Gemini API in Google AI Studio and Antigravity, while Chrome browser support is scheduled for a future release.
Status of Additional Live Voice Models
Google initially indicated that Gemini 3.5 Live and 3.5 Live Experimental models would arrive alongside the transcription update. However, the company clarified that those interactive voice tools are not ready for deployment yet. While Gemini Audio transcription is live today, Google has not provided an updated launch date for the remaining voice models.
