Audio3h ago
Google Launches Gemini 3.5 Transcribe
A 5.5 percent live-speech error rate defines the new Gemini 3.5 Transcribe model. It processes voice-to-text 70 percent faster than the previous Chirp 3 engine. The tool automatically removes filler words and corrections to polish raw audio. Google will integrate the model across its ecosystem after its initial Pixel 11 debut.