Google Launches Gemini 3.5 Transcribe Model
Google has launched Gemini 3.5 Transcribe, a new speech-to-text model designed for faster and more intelligent voice interactions. The company introduced the model on August 26, 2026.
The model already powers Rambler, an AI-assisted dictation feature in Gboard on Android. Meanwhile, Google plans to bring Gemini 3.5 Transcribe to Chrome, extending AI-powered voice input across more everyday applications.
Gemini 3.5 Transcribe Improves AI Dictation
Gemini 3.5 Transcribe builds on Google’s earlier Chirp 3 speech recognition technology. However, the new model goes beyond conventional word-for-word transcription. It can understand natural speech, remove filler words, and produce cleaner, formatted text.
Furthermore, the model can recognize specialized terminology through custom vocabulary. It can also handle language switching during a conversation, which could improve transcription for multilingual users. Google says Gemini 3.5 Transcribe supports more than 85 languages.
The system also handles self-corrections within speech. For example, users can correct themselves while speaking without manually editing the resulting text. In addition, the model supports word-level timestamps and speaker identification for recorded audio.
Google reports substantial performance gains over Chirp 3. According to the company, time to final transcription improves by 70%, while FLEURS testing produced a 5.50% word error rate in streaming use and 5.04% in non-streaming applications.
Rambler Brings Intelligent Voice Editing
On Gboard, Gemini 3.5 Transcribe powers Rambler, which turns spoken thoughts into polished written text. The feature removes filler words and automatically handles grammar and punctuation. Moreover, users can issue natural voice commands to edit or rewrite text.
Rambler can also change the tone or style of an existing draft. Therefore, users can dictate a rough message and then ask the system to make it shorter, more professional or otherwise revise it.
The feature currently requires compatible devices and is available in selected countries and languages. Google says Rambler works across applications where Gboard operates.
Beyond Gboard, Google has integrated Gemini 3.5 Transcribe into its broader ecosystem. Developers can access the model through the Gemini API, Google AI Studio and Google Enterprise Agent Platform. It also supports real-time voice agents, captioning and post-call analytics.
Chrome Expansion Extends Google’s Voice Strategy
Google plans to bring Gemini 3.5 Transcribe to Chrome soon. Once available, users will be able to dictate text directly into web fields, including replies, posts and prompts.
Consequently, the technology could turn voice input into a broader interface for web applications. Instead of simply converting speech into text, the system can interpret intent and apply formatting or edits.
The launch also reflects Google’s wider push to integrate Gemini across everyday computing. With Rambler, Chrome and developer APIs, Gemini 3.5 Transcribe extends from smartphone dictation into productivity and software development workflows.
For developers, the model offers a foundation for voice-driven applications. For consumers, meanwhile, it promises a more natural way to dictate and edit content without having to manually correct every spoken mistake.
