01 — Privacy
Audio never leaves your machine.
Whisper runs locally. Speech is processed in RAM and discarded the moment we have your text. Nothing on disk, nothing on a server. No BAA, no GDPR transfer assessment needed.
Beta · Pro
After your words are transcribed, a local LLM cleans them up — no API key, no cloud call, nothing uploaded. Three things it can do today.
Fixes typos, drops filler words, and restores brand names and code terms — tuned to the kind of text you write.
so umm i pushed the fix to githab and the the latency droped on postgress
So I pushed the fix to GitHub, and the latency dropped on Postgres.
Speak in your native language and get the text in another — translated locally on your computer, the moment you finish.
Hi, I'm sending over the quarterly report — take a look when you get a chance.
Hola, te envío el informe trimestral. Échale un vistazo cuando puedas.
Turn casual speech into the register you need — professional, formal, or technical — without rewriting it yourself.
hey can you fix this asap its kinda urgent sorry
Could you address this when you have a moment? It is fairly time-sensitive.
Workflow
No browser extension, no app integration. Download once, then one hotkey and one paste destination — your cursor.
Download and install SnailText. The model downloads once — then everything runs offline.
Hit the global hotkey. The pill at the bottom of your screen wakes up.
Talk like you would to a colleague. Whisper handles punctuation and silence.
Hit the hotkey again. Words land at your cursor, in whichever app you were in.
Speed
Watch the same prompt land in two ways. The keyboard takes its time. Your voice doesn't.
Press
Speak
Recognise
Paste
Same prompt, two paths. An extra hour back, every working day.
Universality
No plugins, no app-specific integrations. SnailText pastes the same way you would — ⌘V — anywhere your keyboard works.
Built for
Vibe-coders
Send 200-word prompts to Cursor, VS Code, Windsurf in 20 seconds.
Code with voiceDevelopers
Custom dictionary (Pro) picks up Kubernetes, gRPC, async/await.
Ship fasterWriters
Tweets, blog posts, scripts. Audio stays on your device.
Draft out loudStudents
Free on tiny + base models. Works offline in libraries.
Study fasterPMs
Meeting notes, long emails, Slack threads. Three devices, one plan.
Cut the writing taxWhy people pick it
01 — Privacy
Whisper runs locally. Speech is processed in RAM and discarded the moment we have your text. Nothing on disk, nothing on a server. No BAA, no GDPR transfer assessment needed.
02 — Speed
GPU mode via Vulkan (Windows) or Metal (Mac) transcribes a 10-second phrase in under 2 seconds. No GPU? The free Tiny and Base models finish in 1–3 seconds on CPU — fast enough to keep up with your thinking.
03 — Local AI
Pro includes Gemma running locally — no API key, no subscription to OpenAI. It fixes transcription errors and formats code identifiers (snake_case, camelCase) after you speak. Everything stays on your machine.
04 — Cross-platform
Most desktop dictation apps treat Windows as an afterthought. SnailText is built equal for both — same hotkey, same UI, same models. One Pro plan covers three devices.
Pricing
The free tier is unlimited and covers the everyday case. Pro is for the larger local models and the rare edge case where accuracy matters.
Free
Pro
$89 billed yearly
Early access
"I switched from Wispr Flow when I realized I was paying $15/mo to send my company's code review comments to a third party. Same speed locally, plus my work Windows ThinkPad and home M2 Air share one license."
Alex
Backend Engineer · Early access feedback
"I draft long-form on the SnailText pill more than I type now. The trick: I added 40 names to the custom dictionary so it never butchers my characters' names again."
Maya
Freelance Writer · Early access
"I'm not a writer, but I dictate Cursor prompts. SnailText feeds Cursor a 300-word context in 20 seconds. I'd be at 60 seconds typing."
Daniyar
Solo Founder · Early access
FAQ
Unlimited. No watermark, no monthly word limit, no trial timer. You can use the two smallest Whisper models for as long as you want — they cover most everyday dictation.
No. SnailText runs Whisper locally on your machine. Audio is processed in RAM and discarded — nothing gets uploaded, nothing gets logged on a server. You can verify it in your browser/network tab.
Yes. The two free models run on a regular CPU at 1–3 seconds for short phrases. A GPU makes the bigger Pro models faster, not the small ones. We auto-detect Vulkan on Windows and Metal on Mac.
SnailText supports 13 languages out of the box: English, Spanish, German, French, Italian, Portuguese, Dutch, Russian, Polish, Turkish, Chinese, Japanese, and Korean. More can be added on request.
Yes, one click in your account. You drop back to the free tier, your local data stays, and the bigger models become unavailable until you resubscribe.
Go deeper
Stop waiting
No account. About thirty seconds to install. Then a hotkey for the rest of your life.
⌥ Space — that's the only thing you need to remember.