Google 发布 Gemini 3.5 Transcribe，智能转写语音为“实际意图文本”，可消除口误、过滤语气词、格式化数字日期并适配自定义词汇。
Incrdible release from Google, and one more friction gone from ditching my keyboard.
Also looks like, Gemini 3.5 Transcribe will put real pressure on many startups producing clean, well-formatted voice-to-text.
you speak into Gemini 3.5 Transcribe, and instead of directly writing down every word literally, it will produce the text you actually intended.
Its smart mode resolves false starts, strips filler, formats numbers and dates, and biases recognition toward custom vocabulary instead of returning speech word-for-word.
It is available through API right now. There are 2 endpoints:
- gemini-3.5-transcribe: recorded audio, up to 1 hour, with timestamps, speaker identification, custom vocabulary and smart transcription.
- gemini-3.5-transcribe-live: real-time WebSocket streaming, with smart transcription and sub-second interaction latency; individual live sessions currently max out at 10 minutes.
Say hello to Gemini 3.5 Transcribe! - Build apps that understand user speech / intent, even w/ multiple speakers! - Auto-detection of 85+ languages out of the box - Custom vocab adaptation for specialized jargon... SGTM:) API available now in @GoogleAIStudio and Gemini Enterprise, or try it in the Gemini app on macOS or Rambler on Android! More details: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5-transcribe/在 X 查看被引用的帖子
来源：Rohan Paul · x.com