Dictate emails, notes, or messages into any Mac app without sending audio to the cloud.
Speak code and commands into an editor or terminal using spoken punctuation and casing rules.
Dictate seamlessly in a mix of English and Hindi within the same recording.
Build voice snippets that expand a short spoken phrase into a longer saved block of text.
| rahulbhardwaj94/vani | aiduckman/claudeusage_latest_may2026 | arnabau/thermalpulse | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | Swift | Swift | Swift |
| Setup difficulty | moderate | easy | moderate |
| Complexity | 3/5 | 2/5 | 3/5 |
| Audience | general | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires an Apple Silicon Mac, Xcode command line tools, and granting microphone and accessibility permissions.
Vani is a voice dictation app for Mac computers with Apple Silicon chips, meaning newer Macs built around M series processors. You hold down a key, speak, and the words you said appear as clean text wherever you were typing, in any app. Unlike many similar dictation tools, everything happens directly on your own computer. Your voice recording is never sent to a server on the internet, and there is no subscription or account needed. Under the hood, it uses a speech recognition model called Whisper, running through Apple's on-device Neural Engine, to turn your speech into raw text. That raw text then gets cleaned up in two steps: a fast rule-based pass fixes basic punctuation and filler words, and an optional local AI model, run through a tool called Ollama, can further polish the wording. If that second AI step tries to rewrite your meaning instead of just cleaning it up, Vani discards its output and keeps your original words. The final text is then pasted into whatever app you were using. The app supports spoken commands like saying new line or full stop instead of typing them, has a special mode for typing code in editors and terminals that understands things like spoken punctuation for commands, and can switch between English and Hindi within the same sentence. It also keeps a private, searchable history of past dictations, lets you teach it words or names it commonly mishears, and refuses to type into password fields for safety. Setup requires an Apple Silicon Mac running macOS 14 or later, Apple's command line developer tools, and optionally Ollama if you want the AI polish step. It can be installed through Homebrew or built from source using Swift. The project is released under the MIT license. This tool suits Mac users who want a fast, private dictation tool and are comfortable installing developer tools and granting system permissions like microphone and accessibility access.
A fully offline, private voice dictation app for Apple Silicon Macs that types clean text into any app you use.
Mainly Swift. The stack also includes Swift, WhisperKit, Ollama.
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.