Today, we’re introducing Gemini 3.5 Transcribe, our most precise speech-to-text model yet, designed for intelligent voice interactions in Google Antigravity. It pairs screen context and chat history, with your permission, to ensure pinpoint transcription accuracy across file names, agent thoughts, and active documents.

Try it today

Oh my god, I thought I'd gotten a surprise, but it seems like it's completely irrelevant to me. Come on, hurry up and get the big model out!

Demo with 3.6 Flash and not 3.7 Flash

You guys could make a free Wispr Flow killer with this what are you doing

Speech-to-text that understands your screen context could be a game changer.

Can we get remote for antigravity IDE also?

Dictation only works for me if it never eats my variable names. If it can hear "spellWrightRenderer" off screen context and not turn it into "spell right renderer", I'll actually use it.

This not the Gemini 3.5 we want, cmon guys you have to do better

Nice. It is already present in Antigravity for everyone

i thought it's Gemini 3.5 pro 🥺

please just an MCP for Gemini Notebook that's it

i dont use it！I am unable to use the voice input feature in Antigravity 2.0. Even after following the manual, clicking the microphone button causes it to automatically shut off after 2 seconds, preventing it from recording any audio or outputting text.

Everyone at Google is active today

hey google why are you using gemini 3.6 flash instead of your *newest* and most *capable* model which is 3.7 flash?

@rodydavis I was lucky enough to try this feature early. My eyes were wide open when I saw the text adapt to the context. Very impressive 😆

@WisprFlow please check

Is it available here? If not yet, when will it be available here?

Googleantigravityは、proモデルを追加しないのであれば 最先端のオープンモデルを追加すべきです Claude sonnet Opusそして、GPT-OSS120Bを廃止して、 GLM、Kimi、Qwenといったオープンモデルの最新バージョンを並べるべきでしょう Claudeに使用料を払うより、オープンモデルを自社サーバーに設置する方がGoogleの経営的にも良いはずです

Agy cli please

Will gemini 3.5 pro comes or not?

Still don't have voice input on mine! Wondering how accurate it is with real messy audio—mumbling, people cutting each other off, etc. Also noticed the demo uses Gemini 3.6 Flash. Is 3.6 better than 3.7 for this, or was this finished a while ago?😅

sounds nice. I wanna know something else: how accurate is it when ppl talk over each other, mumble, use slang, and drop 12 inside jokes in 20 sec? Transcribing English in a demo ≠ transcribing how ppl actually talk.

It can pick nearly all the accents?

Is it possible to use Gemini 3.5 Transcribe standalone?

我更希望能用antigravity批量地给音频文件做转录，而不是只能用麦克风

Finally, a voice dictation model that actually understands what is happening on my desktop screen. This makes hands-free workflows a whole lot more viable.

Why not open source?

Amazing! Google's speech recognition is like nothing else.

stupid trash, you ashole

Incredible accuracy boost context aware transcription is a game changer!

So whats the pricing like

at what point does context aware transcription become context biased transcription ?

Как вам идея добавить её в GBOARD ??? Почему я до сих пор должен диктовать предложение с ошибками @googledevs сделайте уже наконец-то интеграцию нейросетей в свои продукты

it's cool