Google has launched Gemini 3.5 Transcribe, a speech-to-text model that removes filler words and formats your voice notes as you dictate. It rolls out today to the Gemini app on macOS and to Rambler, the dictation feature in Gboard on Android, with Chrome support to follow.
The audio models arrive while Google’s flagship Gemini 3.5 Pro, promised for June, still has no release date, The Verge notes. Transcribe launches alongside Gemini 3.5 Live and 3.5 Live Experimental, which handle voice chat and step-by-step reasoning respectively.
What Gemini 3.5 Transcribe does
Transcribe replaces Chirp 3, Google’s previous transcription model. According to the official announcement, it averages a word error rate of 4.0 per cent for streaming and 2.6 per cent for pre-recorded audio, per Artificial Analysis, with time to final transcription cut by 70 per cent.
The model handles self-corrections (“let’s meet Tuesday — no, Wednesday”), recognises more than 85 languages, and accepts custom vocabularies so specialised jargon and odd spellings survive the edit. In pre-recorded audio it attributes speech to up to three speakers with word-level timestamps.
Rambler on Android
On Android, the model powers Rambler, the Gboard dictation feature Google announced in May. Rambler turns spoken thoughts into formatted text, strips filler words, and takes voice commands to rewrite or restyle a draft. It supports Arabic among eleven languages, but currently requires a Pixel 11 and is rolling out in select countries and languages, per Google’s Rambler help page — which means most Android phones keep their “ums” for now.
Developers can reach the model in public preview through the Gemini API in AI Studio and Antigravity, and via Gemini Enterprise Agent Platform.
What this means in the UAE
The useful part for UAE users is that Arabic is supported, and the model itself detects over 85 languages including regional accents and dialects. The Gemini app on macOS works in English today, which covers most dictation use cases for journalists and creators.
On Android, the bottleneck is the device: Rambler needs a Pixel 11, and Google has not confirmed a UAE rollout. This is separate from Google’s offline dictation app, which runs on iPhone with no internet connection and transcribes with zero filler words. Chrome support, already shaping up in the browser, is coming to any web field, but Google has given no date.
What is Gemini 3.5 Transcribe?
Google’s latest speech-to-text model. It turns raw audio into clean, formatted text, removing filler words like “um” and “uh”, handling self-corrections, and recognising more than 85 languages. It replaces Google’s previous transcription model, Chirp 3.
When is Gemini 3.5 Transcribe available in the UAE?
The Gemini app on macOS works in English now. On Android, the model powers Rambler in Gboard, which supports Arabic but currently requires a Pixel 11 and is rolling out in select countries and languages. Google has not confirmed a UAE rollout, and Chrome support has no date yet.
How is Gemini 3.5 Transcribe different from standard voice typing?
Standard voice typing transcribes word for word, mistakes included. Transcribe cleans up as it goes: it removes filler words, formats grammar and punctuation, and can attribute speech to up to three speakers with word-level timestamps.
