Google Launches Gemini 3.5 Transcribe with 85+ Language Support

Google introduced Gemini 3.5 Transcribe, a speech-to-text model that auto-formats transcripts, identifies up to three speakers with timestamps, and cuts word error rate to 4.0% streaming / 2.6% pre-recorded with 70% lower latency than its predecessor.