Google announces Gemini 3.5 Transcribe for AI-powered speech-to-text

Google unveiled Gemini 3.5 Transcribe, an AI model that removes filler words and corrections to deliver polished text from voice input, now extending beyond Gboard’s “Rambler” feature on the Pixel 11 to the broader Google ecosystem. The model is roughly 70% faster than the prior Chirp 3 engine and reduces live-speech error rates to 5.5% versus Chirp 3’s 7.32%.