Dictate text on a Mac with a hotkey instead of typing, with filler words and mistakes cleaned up automatically.
Dictate in English, Tagalog, or a mix of both with automatic language detection.
Avoid a subscription dictation service by using your own Deepgram and Groq API keys.
Build the app from source to confirm it ships with no telemetry.
| mrkvn/muni | anonfedora/openpaste | codeaashu/agents-are-thinking | |
|---|---|---|---|
| Stars | 15 | 15 | 15 |
| Language | Rust | Rust | Rust |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 2/5 | 2/5 |
| Audience | general | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires macOS 14+ on Apple Silicon plus your own Deepgram and Groq API keys.
Muni is a dictation app for macOS. The user holds a hotkey, speaks, and releases it, and the cleaned up transcript is pasted straight into whichever app is currently focused. During cleanup, filler words are removed, spoken corrections like saying scratch that are applied, and grammar and punctuation are fixed while trying to preserve how the person actually phrased things. One notable feature is automatic language detection between English, Tagalog, and a mix of the two called Taglish. This detection runs on the device itself by default, without sending audio anywhere, using a small bundled speech model. Based on which language is detected, the audio is routed to different transcription services: English goes to Deepgram, while Tagalog and Taglish go to Groq's speech model. The cleaned up version of the text is then produced by a language model, using a smaller model for short recordings and a larger one for longer ones. Muni requires macOS 14 or later on Apple Silicon hardware. It uses a bring your own key approach, meaning the user needs their own Deepgram and Groq API accounts and keys, which are entered once during setup and stored securely in the macOS Keychain rather than being sent to the app's own servers. The project can be built from source using Rust and a frontend tool called pnpm, and the README notes that builds made from source include no crash reporting or analytics, while the officially distributed app includes optional, toggleable telemetry. The app itself is free to use, with an optional pay what you like support link, and it is released under the MIT license, which allows free use, modification, and distribution as long as the copyright notice is kept.
A free, open source macOS dictation app that cleans up your speech and pastes it, using your own Deepgram and Groq API keys.
Mainly Rust. The stack also includes Rust, Tauri, Deepgram.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.