Google has released Gemini 3.5 Transcribe, a new AI transcription model that automatically removes filler words, detects specialized jargon, and supports more than 85 languages. The model began rolling out on August 26, 2026.
Gemini 3.5 Transcribe is the latest addition to the Gemini Audio family, following the earlier launch of Gemini 3.5 Live Translate. Google describes it as “a major advancement” over its previous transcription model, Chirp 3, citing improvements in multilingual performance and wording error rates.
The model lets users edit text using their voice and can automatically format transcribed content. It strips out filler words such as “um” and “uh” without manual intervention. Users can also supply a customized vocabulary, allowing the model to adapt to unique spelling requirements and industry-specific terminology. For pre-recorded audio, Gemini 3.5 Transcribe can attribute speech to up to three distinct speakers and provide word-level timestamps.
The rollout is initially limited to English. It is available to macOS users through the Gemini app and to Android users via the Rambler dictation feature in select countries and languages. Developers can access the model in public preview through the Gemini API via AI Studio and Antigravity. Google says Chrome support is coming soon.
The launch arrives as Google has yet to release Gemini 3.5 Pro, a model it had previously said would roll out in June. Google also clarified after providing pre-publication information to The Verge that only Gemini 3.5 Transcribe is being announced at this time — two additional Gemini Audio models, 3.5 Live and 3.5 Experimental, that were mentioned earlier are not part of today’s announcement.
Source: The Verge