Free & open source · macOS 14+ · 70 languages

Voice to text,
instantly.

Frespr lives in your menu bar. Hold a key, speak, release — your words appear wherever your cursor is.

— stars on GitHub — downloads
Download for Mac View on GitHub
You'll need a free API key to get started
⚠️
macOS may block the installer — since Frespr isn't signed with a paid Apple Developer certificate, Gatekeeper will show a warning. To open it anyway:

Option A: Right-click Frespr.pkg → Open → click Open in the dialog.
Option B: Go to System Settings → Privacy & Security, scroll down, and click Open Anyway next to the blocked item.
Recording
0:04
Schedule a team sync for Thursday at 3pm
Right ⌥ Release to inject ✦ Clean up

Everything you need

Designed to disappear

🎙

Toggle to record

Press your hotkey, speak, press again. Text at your cursor instantly — no clicking, no app switching.

⚡️

Deepgram Nova-3

Real-time streaming transcription with ~300ms latency. 70+ languages. Continuous word-by-word updates — no gaps.

✨

Post-processing

Clean up, summarize, or apply a custom prompt automatically after every dictation. Optionally powered by Gemini if you set a Gemini API key.

⌥

5 hotkey choices

Right ⌥, Left ⌥, Fn/Globe, Right ⌘, Ctrl+Option. Pick whatever fits your hands.

📋

History & re-inject

Last 20 transcriptions in the menu bar. Click any entry to re-inject it anywhere.

🌐

70 languages

Speak in English, Spanish, French, Japanese, Arabic, Hindi, and 64 more — Deepgram Nova-3 handles them all.

🔤

Key terms

Add names, product names, and jargon so they are transcribed correctly instead of phonetically.

🔒

Your key, your data

Bring your own Deepgram API key. Audio goes directly from your Mac to Deepgram — no third-party servers.

🛠

Open source

AGPL-3.0, no Xcode needed. Build with Swift CLI tools in one command.

Setup

Up in 60 seconds

Getting Started
1
Download and open Frespr.pkg
2
macOS will block the installer — right-click Frespr.pkg → Open → click Open. Or go to System Settings → Privacy & Security → Open Anyway.
3
Add your Deepgram API key in Settings ⌘,
4
Hold Right ⌥ and speak

API Keys

Get your free API keys

🎙
Deepgram API Key
Required

Powers transcription. Deepgram Nova-3 streams your speech to text in real time with ~300ms latency. Audio goes directly from your Mac to Deepgram — Frespr never touches it.

1
Sign up at console.deepgram.com — no credit card required
2
Go to API Keys → Create a new API Key
3
Copy the key, open Frespr → ⌘, → paste into Deepgram API Key
Get Deepgram Key →
Pricing: Free tier includes $200 of credit (~200+ hours of audio). After that, ~$0.0043 per minute — roughly $0.26/hour of speech. Most users never pay anything.
✨
Gemini API Key
Optional

Powers post-processing. After transcription, Gemini can automatically clean up filler words, fix grammar, summarize, or apply a custom prompt you write. Only the text is sent — no audio.

1
Go to aistudio.google.com/app/apikey
2
Click Create API Key and copy it
3
Open Frespr → ⌘, → paste into Gemini API Key → enable post-processing
Get Gemini Key →
Pricing: Gemini Flash has a generous free tier (15 requests/min, 1M tokens/day). For personal dictation use, you will very likely never exceed it.

How it stacks up

Frespr vs. the alternatives

Frespr
Wispr Flow Superwhisper Willow Voice
Price Free forever $18/mo $12/mo From $9/mo
Open source ✓ AGPL-3.0 ✗ Proprietary ✗ Proprietary ✗ Proprietary
Platform macOS 14+ Mac, Windows, iOS Mac, iOS Mac, Windows, iOS
Transcription engine Deepgram Nova-3 (real-time) Proprietary AI Whisper / cloud Cloud ASR
Works in any app ✓ ✓ ✓ ✓
AI post-processing ✓ Clean up / summarize / custom ✓ Limited Limited
Bring your own API key ✓ Required (free tier available) ✗ Vendor-managed ✗ Vendor-managed ✗ Vendor-managed
Audio sent to third-party servers ✗ Audio direct to Deepgram; text to Google if post-processing enabled ✓ Via Wispr servers ✓ Via Superwhisper servers ✓ Via Willow servers
Live streaming overlay ✓ ✓ ✓ Partial
Transcription history ✓ Last 20, re-inject ✓ ✓ Limited
Languages supported ✓ 70 languages ~30+ ~100 (Whisper) ~30+

Common questions

FAQ

No catch. Frespr is completely free and open source (AGPL-3.0 licence). You bring your own Deepgram API key, which has a generous free tier (plenty for personal use). Deepgram charges fractions of a cent per minute of audio beyond that. Frespr itself will never charge you.
Both do push-to-talk voice dictation in any app. The key differences:

Price: Wispr Flow costs ~$18/month. Frespr is free.
Privacy: Wispr Flow routes your audio through their servers. Frespr sends audio directly from your Mac to Deepgram's API — no intermediary.
Source: Frespr is open source; you can audit and build it yourself.
Platform: Wispr Flow also supports Windows and iOS. Frespr is macOS-only for now.
Superwhisper uses OpenAI Whisper (batch processing after you stop speaking). Frespr uses Deepgram Nova-3, which streams transcription while you speak — you see words appear in real time in the overlay. Superwhisper's paid plans start at ~$12/month. Frespr is free. Both are macOS-native and work system-wide.
Willow Voice is a subscription product (from $9/month) that manages its own AI infrastructure. Frespr skips the middleman entirely — your audio goes from your microphone directly to Deepgram's servers via your own API key. Willow supports more platforms (Mac, Windows, iOS, Chrome extension). Frespr currently supports macOS only but is 100% free and open source.
Frespr never touches your audio. The app captures audio locally and sends it directly to Deepgram's API using your API key. No Frespr server exists. You can verify this by reading the source code — it's all in DeepgramService.swift.
No. The pre-built Frespr.pkg installs directly to /Applications. If Gatekeeper shows a warning, run:

xattr -dr com.apple.quarantine /Applications/Frespr.app

To build from source you only need Xcode Command Line Tools (xcode-select --install). No Apple Developer account, no signing certificate, no Xcode IDE required.

The builder

Built by a real person

Salaheldin Saleh
Salaheldin Saleh
Software Engineer · Builder
I built Frespr because I was tired of paying $18/month for voice dictation. Deepgram Nova-3 is incredible — real-time streaming transcription, generous free tier, and no middleman. Frespr is my attempt to make that accessible to everyone, for free.
Buy me a coffee LinkedIn GitHub