Google has launched Gemini 3.5 Transcribe, a new transcription model designed for highly accurate, context-aware speech-to-text in over 85 languages. The model automatically filters filler words, formats unstructured speech, and uses multimodal capabilities to execute voice commands, now integrated into the Gemini app, Gboard, and the Gemini API.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
Google releases Gemini 3.5 Transcribe, supporting 85+ languages, automatically filtering filler words, executing voice commands with screen context, and now available across multiple platforms.
On August 26, Google AI announced on X the release of Gemini 3.5 Transcribe, calling it the 'latest transcription model' built for precise, intelligent dictation across apps and devices. The company emphasizes that this is not just dictation but 'proactive intelligence,' offering context-aware speech-to-text capabilities.
According to the official announcement, the model supports 85+ languages, automatically filters filler words (such as 'um' and 'uh'), formats unstructured speech, and uses multimodal capabilities to execute voice commands with screen context. In a demo, the model seamlessly transformed messy voice input and local files into a polished email draft.
Currently, users can experience the model in the Gemini app on macOS and Gboard on Android. Developers can build applications using the Gemini API through Google AI Studio, Antigravity, and the Gemini Enterprise Agent Platform (public preview). The company also stated that it will soon be available on Chrome and Gemini Enterprise for Customer Experience.
In replies on X, user Autobunnyhop commented: "'Um... impressive' is exactly the phrase this model would delete before you finish saying it. 85+ languages, and it understands my accent better than most people." While this is a personal experience, it reflects the model's performance in accent recognition.
This report is based on Google AI's official X post, which is a primary source, but the feature descriptions and availability are official claims and have not been independently verified. User comments represent personal opinions and do not constitute objective evidence.
Gemini 3.5 Transcribe marks a shift in speech-to-text from a passive tool to proactive intelligence. Its multilingual support and screen context understanding could reshape human-computer interaction, but its actual performance awaits broad user validation.
Primary report
Primary source
Also covered by 2 sources