Whispur
Now in public beta · v0.3.0

Voice to text,
anywhere on macOS.

Hold a shortcut, speak, let go. Whispur transcribes and pastes into any app — with your choice of speech engine and optional AI cleanup. Open source. Bring your own keys.

Download for macOS View on GitHub

Free · MIT · macOS 14+ · Apple Silicon & Intel

Features

Built like a native Mac app

Global hotkey dictation

Bind a hold-to-talk or toggle shortcut. Press Fn, speak, let go — transcribed text lands wherever your cursor is.

Multiple speech engines

Apple on-device, OpenAI Whisper, Deepgram Nova, ElevenLabs Scribe. Swap providers per workflow.

Optional AI cleanup

Route the transcript through OpenAI, Anthropic, or Groq with your own prompt. Fix punctuation, tone, or format on the fly.

Bring your own key

No subscription, no middleman. Your API keys live in Keychain. Apple STT works free, on-device, out of the box.

Privacy-first

Audio stays local by default. Network calls only happen when you pick a cloud provider — and you can see every request.

Always up to date

Runs quietly in the menu bar. Signed, notarized, and auto-updates via Sparkle — no install friction.

Bring your own key

Your choice of provider.

Pick the engine that fits your workflow — on-device for privacy, cloud for accuracy. Keys are stored in macOS Keychain, never sent anywhere but to the provider you picked.

Speech to text

  • Apple — on-device, free, default
  • OpenAI — Whisper & gpt-4o-transcribe
  • Deepgram — Nova-3 real-time
  • ElevenLabs — Scribe v2

Optional LLM cleanup

  • OpenAI — GPT-4o & GPT-4o-mini
  • Anthropic — Claude Sonnet & Haiku
  • Groq — Llama & Mixtral on LPUs
  • Custom prompts — write your own
A look inside

Designed for your menu bar.

Whispur menu bar dropdown showing recording state and last transcript
Menu bar status with last transcript and quick actions.
Compact recording overlay with waveform and trigger mode
Compact top-center recording surface with live waveform.
Whispur Setup tab with permission checklist
Setup checklist walks you through permissions and provider config.
Provider settings with speech and cleanup provider pickers
Swap speech or cleanup provider in a click — keys live in Keychain.
Install

Running in under a minute.

01

Download the DMG

Signed and notarized by Apple. Double-click to open — no right-click workaround needed.

02

Drag to Applications

Launch Whispur. A small icon appears in your menu bar — that's home.

03

Grant permissions

Microphone for capture, Accessibility for the global shortcut and paste-back. Then hold your hotkey and talk.

FAQ

Good questions.

Is Whispur really free?

Yes. The app is MIT-licensed and free forever. Apple speech-to-text works on-device with no account. Cloud providers bill you directly for their own usage — we never touch your keys or your audio.

Do you see my audio or transcripts?

No. Whispur has no backend. Audio is processed locally or sent directly to whichever provider you configure. Everything stays between your Mac and that provider.

Which macOS versions are supported?

macOS 14 (Sonoma) or later, Apple Silicon or Intel.

Can I customize the shortcut?

Yep. Both the hold-to-talk and toggle shortcuts are rebindable in Settings. Defaults use Fn since it's rarely bound elsewhere.

How does AI cleanup work?

After transcription, Whispur can optionally run the text through an LLM with a prompt preset (fix punctuation, format as email, summarize, etc.) before pasting. Optional per preset.

Where's the source?

On GitHub. Issues and PRs welcome.

Ready to stop typing?

Download Whispur, grant two permissions, and dictate into anything in the next 60 seconds.

Download for macOS