Dictation for people who think out loud
Speak like you think.
Read like you wrote it.
Plainspoke strips the ums, the false starts and the backtracking, and leaves your meaning, your hedges and your emphasis exactly where you put them. It is not a summarizer — and it shows you every word it touched.
No credit card. The free tier is not a trial clock.
okay so, um, I need to— I think I need to add rate limiting to the API, uh, the the login endpoint mostly, because like we're getting hammered. Maybe also the signup one? I'm not sure about that one actually. But definitely login. And it should probably be per IP, um, per IP address I mean, not per user, because obviously they don't have a user yet when they're logging in. Sorry, when they're failing to log in.
I think I need to add rate limiting to the API, mostly on the login endpoint, because we're getting hammered. Maybe also the signup endpoint — I'm not sure about that one. But definitely login.
It should probably be per IP address rather than per user, since they don't have a user yet when they're failing to log in.
Highlighted: every hedge and emphasis that survived the rewrite.
“Add per-IP rate limiting to the login and signup endpoints.”
Shorter, cleaner, and it decided the signup question for you — the one thing you said you weren’t sure about. Your agent now builds it. That is the failure mode Plainspoke exists to avoid.
Transcription was never the hard part.
Speech-to-text is solved and nearly free. What nobody hands you is the step after: the thing you said, turned into something you can actually send, paste or ship — with the reasoning intact.
So people either type instead, or dictate and then spend as long repairing the transcript as they saved.
“I would be keen to use voice to control Claude Code, but I have always felt that the way I speak is different from the way I write good prompts.”
Three things Plainspoke will not do to your words.
It will not summarize you.
Summarizers optimize for shorter. Plainspoke optimizes for the same thing you said, written properly. If you spent four sentences circling a point, you get four sentences — without the ums.
It will not decide what you meant.
Every “I think”, “maybe” and “most importantly” is load-bearing. A model that flattens your hedges into confident statements hands your agent a spec you never approved.
It will not hide what it changed.
Every transcript keeps its verbatim original. See the diff, take any edit back, or revert the whole thing to exactly what came out of your mouth.
Every device, one history
Record on your phone. Paste on your laptop.
Start talking on whatever is in your hand. The transcript lands on every device you are signed in on, live, while you are still speaking — then waits in your history until you sit down.
Most dictation tools store transcripts on the device that recorded them, which means the thought you had on a walk is stuck on your phone.
Open source client
Built on Linux. Not ported to it.
A global hotkey, a tray icon, and cleaned text on your clipboard before you have finished reaching for it. No Wayland caveats buried in an issue tracker.
The desktop client is open source under the MIT licence. A tool that listens to your microphone should let you read what it does with it — and let you run it without us, forever.
$ plainspoke
listening on Super+H · tray icon active
● recording… 0:11
cleaned · copied to clipboard
Priced so the free tier is worth using.
Every plan includes the cleanup and the verbatim original. Paid plans buy you more minutes, live preview, and history that never expires.
Free
Enough to actually use, not an eight-minute trial.
No card, no trial clock
Start free- 120 minutes of cloud transcription a month
- Intent-preserving cleanup on every transcript
- See exactly what changed, revert to verbatim
- Linux desktop app, unlimited local transcription
- 7 days of history
- No live preview while you speak
- No multi-device sync
Pro
Most peopleFor people who dictate most of what they write.
or $96 a year — 2 months free
Get Pro- 1,500 minutes of cloud transcription a month
- 300 minutes of live preview while you speak
- Multi-device sync — record on your phone, paste on your laptop
- History kept forever, searchable
- Everything in Free
Max
For all-day dictation and long sessions.
or $192 a year — 2 months free
Get Max- 5,000 minutes of cloud transcription a month
- 1,000 minutes of live preview
- Priority processing
- Everything in Pro
Students and open-source maintainers get 50% off — email us.
Questions
- How is this different from Wispr Flow or Superwhisper?
- Two things. They keep transcripts on the device that recorded them — Plainspoke syncs your history across every device you sign in on, so you can capture on your phone and paste on your laptop. And they clean up your speech without showing you what they removed; Plainspoke keeps the verbatim original next to the result, always.
- Does it work on Linux?
- It was built on Linux first. The desktop app has a global hotkey, a tray icon, and copies the cleaned text straight to your clipboard. It is open source, so you can read exactly what it does with your microphone.
- What about macOS and Windows?
- The web app works on any device with a browser, including phones. Native macOS and Windows clients are not built yet, and we would rather say so than imply otherwise.
- Do I have to pay to get anything useful?
- No. Free gives you 120 minutes of cloud transcription every month with the full cleanup, plus unlimited local transcription in the Linux app. Paid plans add live preview while you speak, multi-device sync, and history that is kept forever.
- Is my audio kept?
- Audio is processed and discarded — only the transcript is stored, and only against your account. Transcripts are protected by row-level security in the database, so no other account can read them.
- What counts as a minute?
- Audio you actually record, rounded to the second, not wall-clock time with the app open. Sessions are capped at 10 minutes each so a forgotten open microphone cannot drain your month.