AI Generated Image
Google launches Gemini 3.5 Transcribe, which powers Rambler
9to5Google • • technology
Key Points:
- Google launched Gemini 3.5 Transcribe, its most accurate speech-to-text model, which is already integrated into several Google products like Gemini Live and Gboard Rambler.
- The model excels in noisy environments, handles jargon, self-corrections, filler words, and formats text naturally, achieving a low Word Error Rate (WER) of 4.0% for streaming and 2.6% for non-streaming transcription.
- It supports over 85 languages with automatic detection, recognizes custom vocabulary, and can identify multiple speakers with timestamps for up to three speakers.
- Gemini 3.5 Transcribe improves transcription speed by 70% compared to the previous Chirp 3 model and enables voice-driven task execution by delegating complex commands to other Gemini models.
- The technology is currently available on macOS, Android, and Google Antigravity’s prompt box, with plans to launch on Chrome browser for voice dictation across web fields.
