About
Voice is becoming the way we work with computers. We're building the layer that makes it predictable.
Why we exist
Open your laptop today and count the microphone icons. Google search has one. ChatGPT has one. Slack has one. Your phone keyboard, Cursor, Notion AI, the new tab in Edge — all of them have a little mic.
None of them work the same way.
One records for 30 seconds, then cuts you off. One only understands English. One mangles the names of your colleagues. One sends your audio to a server you've never heard of. One has no shortcut, so you have to find it with the cursor every time. You learn each one by trial and error, and a week later you're guessing again.
SnailText is the opposite bet: one hotkey, one behaviour, in every text field on your computer. You learn it once. You know what it can do, what it can't, and where your audio goes. Then you use it everywhere — search bar, IDE, email, AI chat, Slack thread, the comment box on Reddit. Same press, same speak, same paste.
The shift
Two shifts are happening at the same time, and they're bigger than dictation.
For decades the keyboard tied you to a desk. Voice doesn't. You can stand up, walk around the room, lean back, pace while you think. People draft long messages while making coffee. They dictate prompts on a stationary bike. They keep working through a wrist injury. None of that is possible with two hands on a keyboard.
Five years ago you had to type a clean prompt for an AI to give you a useful answer. That's no longer true. Modern models — Claude, ChatGPT, Cursor agents, Copilot — handle messy, half-formed, stream-of-consciousness input. You can ramble for ninety seconds about a bug you're chasing, and the model will pull the actual question out of it. You don't have to structure your thoughts before you speak them. You don't even have to finish sentences.
The cognitive tax of "phrase this properly first" is just gone. SnailText is the input layer for that world.
What we believe
01
You shouldn't have to learn a new dictation widget every time you switch apps. Same hotkey, same behaviour, same model, same language settings — across your browser, your IDE, your AI chat, your Slack. One thing to trust, then forget about.
02
Speech is data. Most voice tools treat it as theirs — to log, train on, or pipe to a "context-aware AI." Ours stays in RAM, gets transcribed, and disappears.
03
A 15-minute trial isn't free, it's a countdown. A 2,000-word weekly cap isn't free, it's a meter. SnailText's free tier is unlimited on the two smaller Whisper models — for as long as you keep the app installed.
04
We charge a subscription because notarisation, code signing, and ongoing model updates aren't free. We don't charge it because we want to ration features. Pro unlocks bigger models. The free tier doesn't lose anything.
Commitments
Not in Free. Not in Pro. If we ever add an opt-in cloud STT for Pro users who want it, the toggle will be off by default and labelled clearly.
What you said is what gets pasted. AI rewrites — if we ship them — will be opt-in per mode, never the default.
We don't have your audio. We can't sell what we don't have.
One Pro subscription covers up to three of your devices. Your laptop, your home box, your spare — same plan.
Your dictation history lives in a SQLite file on your machine. Cancel Pro, uninstall the app — your text is still there.
The team
A small team that types too much. We've built developer tools, infrastructure software, and the occasional indie product before. We use SnailText every day — for code comments, Slack drafts, AI prompts, and yes, this About page.
We're funded by the Pro tier, not by investors asking us to ship a meeting transcription feature next month. That's slower, but it means we work for the people using the product, not for a board deck.
Built in Asturias, Spain. Shipped to whoever needs it.
Roadmap
We try not to over-promise. Rough order after the Windows MVP:
Same UX, same models, same hotkey. We're in bring-up.
German, French, Spanish, Portuguese. Translation pipeline ready, locales are next.
For users who want top-shelf accuracy without local hardware. Strictly opt-in, off by default.
Never automatic. If you want a "fix grammar" pass, you turn it on for a specific mode.
We won't build file transcription. MacWhisper does that well; we don't need to be every voice tool.
Try it
Free to start, no account needed. About thirty seconds to install.
