Vox thinks with you. And yes, it transcribes too.

Today, in English and Portuguese — screens, menus and error messages built for both from day one, not translated afterwards.

VOX uses your own access key to a transcription service — Groq comes pre-selected, with generous free use.

It runs Whisper, the engine most people consider the best audio transcriber out there — Groq just delivers it about 5x faster than the original, thanks to their own hardware, without losing accuracy.

That's why, with your own key, your data never passes through our servers: the audio goes straight from your browser to Groq, no stop in between. To try it you don't even need one: your first transcriptions and lenses are on us. Only those go through a bridge of ours to Groq, and the bridge keeps neither the audio nor the text.

Yes — and it's not a trial with a deadline. Free is genuinely free, no card, no sign-up. On top of that, when you start you get 14 days of Pro as a gift: everything unlocked, no card. When it ends you simply stay on Free, and nothing you made is deleted. People who subscribe to Pro (from US$6/month) usually do it for the extra features.

MP3, MP4, WAV, M4A, OGG, WebM, FLAC, AMR, WMA, 3GP — audio or video, pretty much any common format. For video, VOX just pulls the audio out to transcribe it.

Yes — and not just WhatsApp. It also works with screen recordings, on phone or computer: VOX pulls the audio out of the video and transcribes it fast.

Yes. There's an attachment just for that: drop any audio or video — a meeting recording, a lecture, an interview recorded in another app — and VOX transcribes it the same way.

No. VOX runs straight in your browser — nothing to install, no account to create. You just open it and use it; the only setup is pasting in your own API key to switch transcription on.

Free is meant for getting to know VOX, with usage limits in place. VOX Pro is built for everyday use and unlocks wider access to the platform's features, subject to fair-use policy and any technical limits that apply.

You choose: US$6 a month, US$59 a year (works out to US$4.92/month), or a US$149 one-time lifetime payment. You can switch billing cycles whenever you like.

Yes. It's called diarization — automatically splitting out who said what, instead of handing back everything mixed together. In a consultation or interview, that's the difference between a text you still have to untangle and one that's already ready: "Doctor: ... / Patient: ...".

VOX uses AssemblyAI for this — the tech most people consider the most accurate on the market today, with about US$50 in free credit to start (good for hundreds of hours). After that, it runs about 15-20 cents per hour, plus 2 cents for speaker separation.

One thing worth deciding with your eyes open: on the free credit, AssemblyAI's terms allow them to train their models on your audio — only paying customers can turn that off. For a confidential consultation or interview, it's worth weighing before you choose. VOX shows this information (the "confidentiality badge") on every provider, so you decide with what you know, not what we promise.

You can — and there's a way not to even depend on that.

With your own key, there is no Vox server in the path: the audio goes straight from your browser to whichever transcription service you pick, and each one treats data differently. That's why every option carries a confidentiality badge, shown before you choose.

Now, when the matter is truly sensitive, there's the mode that closes everything off: the local transcriber, for Windows and Mac. You install it once and Whisper starts running inside your own computer — the audio goes nowhere, there is no internet in the path. On an ordinary machine, with no graphics card, the Fast model takes about 18 seconds per minute of audio.

The text AI (summary, minutes) runs on your machine too, if you want: just point VOX at Ollama or LM Studio.

One limit worth knowing up front: separating who said what doesn't run locally yet — in this mode the transcript comes out as running text.

And whatever gets stored, you lock: any world becomes a vault with a password, encrypted on the device. Transcribe with no internet, store under key — the data never leaves your desk.

It is. If you talk to AI a lot, you've probably hit this: words coming out wrong in the transcript, whatever app you use most failing or going down right when you need it, the free plan's message limit, and no easy way to hold on to what you already said.

VOX saves every prompt in a project folder, nothing gets deleted — and lets you pick, from several transcription engines, whichever one understands your voice best.

Fluid thinking, documented: you talk, the text gets saved and organized, ready to become your next prompt.