3.6 v3.6.1 · SwiftUI Preferences, Liquid-Glass recording panel

Dictate anywhere.
Polished text appears at your cursor.

Press Option+Space, speak, stop. Whisper Voice transcribes, reformats for the target app (Slack, Mail, Markdown…), and pastes — all in one breath.

Free forever Open source (MIT) 100% local option BYOK (OpenAI / Mistral) macOS 13+

Download not working? Browse all versions

12 seconds. Slack. No tricks.

Real dictation in Slack. Watch the auto-mode detect the app, the panel float in, the waveform breathe with my voice, and the text paste back with inline-code rendering — the kind of thing you'd normally need to manually toggle Slack's markup setting for.

Select text. Speak your edit. Done.

Voice-controlled Cursor for any text field. Highlight an email, speak *"make it shorter and less formal"*, and the edit is pasted back in place.

The voice Cursor that works everywhere.

Your cursor becomes a voice-operated editor. No extension, no sidebar, no context switch. Works in Gmail, Notion, Slack, Linear, even plain textareas.

  • Keeps the file/URL context as a hint for the LLM
  • Git repo, branch and cwd auto-injected when you're in a terminal or IDE
  • No text selected? Speak a free-form instruction — it acts like a pocket assistant
1 · Selected text
"Hi team, I want to let you know we are launching the new feature next week…"
2 · Spoken instruction
⏺ "Make it one punchy sentence with one emoji."
3 · Pasted back
🚀 New feature ships next week — team, hold tight.

Built for the way you actually write.

6 processing modes

Brut, Clean, Formel, Casual, Markdown, Super. Shift-cycle during recording. Or create your own with custom prompts.

Auto-mode by app

Map Slack → Slack mode, Mail → Formel. Whisper Voice detects the frontmost app and switches automatically. Resets to Brut between unrelated apps.

Projects with smart tagging

History auto-tags entries to your projects via git remote, workspace hint, browser host. Filter, search, retro-tag in bulk.

Full pipeline audit

Every dictation logs input, system prompt, output, char delta. Built-in log viewer. You always know exactly what the LLM did with your voice.

Dual-format clipboard

HTML + plain text. Slack, Notion, Gmail get rendered bold/code/lists. Terminal and editors get the raw markdown. No <br> nonsense.

Custom vocabulary

Add your product names (PostHog, Chatwoot…) once. Biases Whisper AND the LLM prompt so your jargon survives transcription and reformatting.

vs. the paid alternatives.

Whisper Voice vs Superwhisper, FreeFlow / Whispering, and macOS dictation. No marketing lies — the honest rundown.

Whisper Voice Superwhisper FreeFlow / Whispering macOS Dictation
Price Free forever €10/month Free (BYOK) Free
Open source Yes · MIT No Yes No
BYOK (your API key) OpenAI · Mistral · local Bundled Yes
100% offline option whisper.cpp built-in Yes Partial Yes
AI post-processing modes 6 built-in + custom prompts Yes Limited No
Super Mode (voice-edit selection) Yes No No No
App-aware auto-select Yes Partial No No
Project tagging + history Yes · auto-tagged No No No
HTML clipboard (Slack markup) Yes No No No

Paid €10/month for what a weekend could ship.

I was paying €10/month to Superwhisper. Great app, but I used it 5 minutes a day and I knew the tech was an OpenAI Whisper API call behind a UI.

So I ran the numbers: €120/year × thousands of users = a fat margin on top of a ~$0.06/hour API cost. I figured the UI was the only real moat.

Three weekends later, Whisper Voice had feature parity + a few things Superwhisper doesn't have (Super Mode, project auto-tagging, dual-clipboard). I open-sourced it under MIT.

If you dictate, I want this to be the last voice-to-text tool you ever install. If you want to fork, rebrand, or ship your own flavor — go for it.

Stop paying for an API wrapper.

4 MB. Free forever. Your keys, your data, your Mac.

v3.7.0 · Signed + notarized by Apple · macOS 13+ · Source on GitHub
Download not working? Browse all versions