Google Launches Gemini 3.5 Transcribe

Google just released Gemini 3.5 Transcribe, a new tool that turns spoken words into text. It works in over 85 languages and is much more accurate than previous versions. The model is also 70 percent faster at finishing tasks than the older Chirp 3 system.

You can now choose between two ways to use it. The streaming version works instantly for live conversations but skips details like speaker names. The batch version takes a little longer but includes full speaker labels and costs half as much.

This split helps developers build better voice apps. If you need quick responses for a bot, use the streaming mode. If you need a full record of a meeting with clear speakers, the batch mode is the better choice.