Google Unveils Gemini 3.5 Transcribe for AI Speech-to-Text

Google launched Gemini 3.5 Transcribe on August 27, 2026, an AI model that reduces live-speech error rates to 5.5 percent while accelerating processing speeds by approximately 70 percent.

The new model supports 85 languages and automatically removes filler words like 'um' and 'uh' to produce polished, ready-to-publish transcripts in real-time.

Currently powering the Gboard 'Rambler' feature on the Pixel 11, the system is being integrated across Google's ecosystem to assist professionals in legal, medical, and journalistic fields.