Skip to content

Local dictation for Apple silicon

Speak. Done.

Turn your voice into clean, useful text anywhere on your Mac. Fast, private, and entirely local.

  • No account
  • No subscription
  • No dictation upload

OnWhispr

Ready · Local
Clean dictationHold ⌃⌥Space to speak
⌃⌥
MacBook microphone
CleanRawNotesEmailMeetingCustom
hey can you send the revised proposal today and ask for feedback by friday
OnWhispr

Hi, could you send the revised proposal today and ask for feedback by Friday?

Inserted0.8 s
01

Stays on your Mac

Speech recognition and optional text cleanup run locally. Your dictation is not sent to an OnWhispr server.

02

Works where you type

Hold one shortcut, speak, and release. OnWhispr inserts the result into the active text field or copies it safely.

03

Pay once, not forever

OnWhispr is being built as a one-time purchase. Final launch pricing will be announced before release.

One gesture

Thought to text, without the detour.

The app lives quietly in the menu bar until you need it.

01

Hold

Press and hold your chosen shortcut.

02

Speak

Use the selected microphone. The live trace confirms that OnWhispr can hear you.

03

Release

Local models transcribe, optionally clean up, and deliver the result.

Private by architecture

Your voice is not a cloud feature.

OnWhispr uses WhisperKit and local GGUF models on your Mac. There is no account, analytics SDK, or dictation upload. After the models are installed, the core dictation flow works offline.

  • 0 dictation uploads
  • 0 analytics SDKs
  • Local history is optional
  • Microphone choice stays app-specific

Model downloads require internet once, or can be provided manually. The recommended setup is about 5 GB; the smaller speech model is about 464 MB.

APPLE SILICON
Entirely on your MacWhisperKit · GGUF · Metal
5 GB recommended models6 output modes0 B dictated data uploaded
Useful output

Not just a transcript.

Choose the shape before you speak. Raw transcription and lightweight Clean dictation work without a formatting model; richer modes use an optional local LLM.

01

Raw transcript

Your words, untouched.

02

Clean dictation

Punctuation, casing, and readable paragraphs.

03

Structured notes

A summary, bullets, actions, and open questions.

04

Email draft

A clear email with an appropriate structure.

05

Meeting notes

Decisions, risks, and follow-ups.

06

Custom prompt

Your template and your rules.

Private beta

Help make dictation feel native on the Mac.

Join the list for beta access and launch updates. No newsletter treadmill, just product news that matters.

Good questions.

Which Macs are supported?+

The current target is Apple silicon (M1 or newer) running macOS 14 or later.

Does dictation work offline?+

Yes, after the selected models have been installed. Downloads need internet once unless you provide compatible model files manually.

How large are the models?+

Whisper Large v3 Turbo is approximately 3 GB and the recommended Qwen formatting model is approximately 2 GB. Whisper Small is about 464 MB and the formatting model is optional.

Can I choose my microphone?+

Yes. OnWhispr remembers an app-specific input device and does not change your Mac's system-wide default microphone.

Is history encrypted?+

History is optional and stored locally, but it is not currently encrypted separately at the application level. You can disable or clear it at any time.

Is there a subscription?+

No subscription is planned. Final one-time launch pricing will be announced before release.