Your hardware is enough

The laptop you already own can transcribe an hour of audio with a state-of-the-art speech model. It can upscale a photo, pull a voice out of background noise, translate speech, lift text off a scan. The models are open, the runtimes are open, and your browser can reach your GPU. So why does almost every AI tool on the web still make you upload your file to their server, ration you three free minutes, and then ask for your card?

That setup made sense when inference needed a data center. It doesn't anymore. For a growing pile of everyday tasks the computer on your desk is more than enough — so the minute caps and the monthly fee aren't paying for compute. They're rent. Software you could just own, rented back to you forever, on the bet that you won't notice the cost went away.

We noticed. YourDevice is what we did about it.

And it's fairer. A tool that runs on your own hardware costs the same for a newsroom and for a broke student: nothing. A transcript shouldn't be a budgeting decision. The same capability for anyone with a browser — not just for whoever can absorb one more subscription.

No server, no catch

You're probably looking for the catch. There isn't one. Every tool here works the same way: the model downloads into your browser once, then runs on your machine — your GPU if you've got one, WebAssembly if you don't. There's no upload step because there's no server doing the work. Your files are never uploaded — everything runs on your device.

Two things fall out of that. One: our costs don't move whether you transcribe one memo or a decade of podcasts, so we've got no reason to meter you — "unlimited" isn't a pricing tier, it's just what's left when nobody's paying per minute.

Two — and this is the one that matters — privacy stops being a promise you have to trust. It isn't a setting we could quietly flip off in next week's update; it's the architecture, and you can check it yourself: open the network tab, or turn your wifi off after a tool loads and watch it keep working. Everyone else asks you to trust them with your files. We never ask for them. Nothing to leak, sell, subpoena, or train on — because your files never left your machine. You don't have to take our word for it. That's the whole point.

Open where it counts

The engine behind all of this is open source, MIT-licensed. The speech model is OpenAI's Whisper — also open. Read the code, run it yourself, or fork it into something better; local-first AI only gets stronger the more people ship it. The tools are free and stay free. No ads today; if they ever land they'll be unobtrusive and tracking-free, and there's a Ko-fi if you want to throw us a coffee. None of it meters the tools or touches your files.

Where this goes

Transcription was first. It wasn't the last — since then we've pushed image upscaling, background and noise removal, OCR, and PDF work across the same line: from metered service to something your own device just does. Hardware keeps getting faster and open models keep getting better, so the line keeps moving, and we'll keep dragging tools across it. Same deal every time — free, unlimited, no accounts, your files never uploaded.

No accounts. No caps. No uploads. AI that runs on your hardware — because it's yours.