Your voice never leaves your Mac.
Free public beta · Apple Silicon · 8 GB RAM · macOS 13 Ventura or later
See it
Hold Fn, speak, release. Clean text lands where your cursor is.
What it is
On-device Whisper, sized to your Mac's memory. Audio never touches the disk.
Text lands at your cursor. Notes, Notion, Slack, VS Code, Mail.
Vocabulary packs tuned for code-switched speech, not a generic toggle.
How it works
Hold Fn and speak.
On-device Whisper.
Filler and punctuation.
Text at your cursor.
Features
Raw, Clean, or Refine ✨ via a local Mistral model. No API keys, no per-token cost. English for now.
Teach it the names it keeps missing, seeded so they come out right.
A short trigger expands to full text. Exact-match, no guessing.
Think out loud in passes. Review before anything is injected.
A local seven-day chart of words dictated and time saved.
Push-to-talk on Fn, right Command, or right Option. Your choice.
It stops listening when you stop talking, not on a fixed timer.
Privacy
Most apps ask you to trust a privacy policy. Ploca removes the need for trust. No code writes your audio to disk or sends your transcripts anywhere. Telemetry was never built.
Compare
Wispr Flow and Superwhisper are good tools with different trade-offs. Here is where Ploca stands apart, and where it does less.
| ploca | Wispr Flow | Superwhisper | |
|---|---|---|---|
| Runs fully on-device | |||
| No cloud path at all | |||
| Works offline | |||
| Local refine, no API keys | |||
| No screen or keylogger access | |||
| Tuned for Tanglish and Hinglish | |||
| Windows and mobile too | |||
| Price | Free in beta | About $15 / mo | From $8.49 / mo |
Competitor details are drawn from their public pages in mid-2026 and may change. Check their sites for current terms.
Who it's for
Why it exists
A pen's job is to write. Not to remember. Not to transmit.
A voice-to-text app should act like a pen. It should hear you and write it down, and nothing more. There are already plenty of notepads, audio archives, and second brains. Those are different products, for different jobs.
I kept looking for that pen and could not find one. I am a writer who is lazy to type, and I do not want a model generating my first draft. Every tool I tried had a trade-off I could not live with. The transcription was not good enough, the privacy was someone else's policy document, or my words were quietly training a model I would never see.
So I built Ploca for myself. Since then I have been publishing at least one essay a week, thinking out loud with Claude instead of typing, and getting thoughts down as fast as they form. The friction is gone.
If that sounds useful to you, I would like to have you in the beta.
Aasif Iqbal J. · maker of ploca
Read the five principles behind ploca →Public beta
Open and free until 30 August 2026, for as long as the beta runs.
We email you the link. Apple Silicon · 8 GB RAM (16 GB recommended) · macOS 13 Ventura or later. The app is about 180 MB; the first launch downloads a speech model sized to your Mac (around 1 to 3 GB), once.
Enter your email and we will send the download link, plus the occasional beta update. We use it for nothing else.
Apple Silicon, 8 GB RAM (16 GB recommended), macOS 13 Ventura or later. No tracking, no third parties.