Google Launches 3.5 Transcribe and Delays 3.5 Live/Experimental

⚡

Key Takeaways

13.5 Transcribe is being rolled out in English on the Gemini app for macOS and via Rambler dictation on Android, with a public preview for developers.

2The model can format text, remove "um/uh," handle up to 3 speakers, and provide word-level timestamps, with custom vocabulary.

3Google is postponing 3.5 Live and Live Experimental without a new date and indicates that only 3.5 Transcribe is announced today.

💡Why it matters — Eligible users now have access to 3.5 Transcribe, while Live updates will be delayed and 3.5 Pro is not yet available.

Google has updated Gemini Audio and introduced 3.5 Transcribe, a transcription feature touted as more robust than Chirp 3. The rollout begins in English on the Gemini app for macOS and via Rambler dictation on Android, with a public preview for developers. Meanwhile, updates for 3.5 Live and 3.5 Live Experimental have been postponed for now, and 3.5 Pro remains anticipated.

Google Delays 3.5 Live and Focuses Announcement on Transcribe

Google has stated that 3.5 Live and 3.5 Live Experimental will not be launched at this time and has not provided a new date. The company further clarifies that only the 3.5 Transcribe feature is being announced today, despite initial plans for updates to 3.5 Live and 3.5 Live Experimental to be available today. The release of Gemini 3.5 Pro, promised for June, is still awaited. Google also announces that support for Chrome for 3.5 Transcribe is expected "soon."

Rollout: macOS, Android Rambler, and Public Preview API

Gemini 3.5 Transcribe is starting to roll out in English for all users of the Gemini app on macOS. The feature is also being deployed for Rambler dictation on Android in certain countries and languages. Developers can access it in public preview via the Gemini API in AI Studio and Antigravity. 3.5 Transcribe represents a new component of the Gemini Audio offering.

⚡Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Cleaner Transcriptions: Formatting, Filters, and Speaker Tracking

The model can automatically format text and remove filler words like "um" and "uh." It assigns speech to up to three speakers in pre-recorded audio and provides word-level timestamps. Users can supply a custom vocabulary, allowing it to adapt to specific spellings and jargon without manual corrections, and to modify naturally with just the voice. Google states that these capabilities cover over 85 languages and specialized jargon, presenting 3.5 Transcribe as a significant advancement over Chirp 3, particularly in multilingual performance and formulation error rates.

What Google Attributes to 3.5 Live and Its Experimental Version

According to previously shared information from Google, 3.5 Live would offer improved performance for managing interruptions within sentences, language detection, and real-time visual processing. The 3.5 Live Experimental variant would go further by narrating its step-by-step process in real-time while tackling more complex tasks. 3.5 Transcribe arrives following the launch of 3.5 Live Translate.