Skip to content

dmclark/VoiceInk

 
 

Repository files navigation

VoiceInk for Voiceitt (Experimental Fork)

What is VoiceInk?

A native macOS app (Swift / SwiftUI, requires macOS 14.4+) that transcribes speech to text and pastes it at the cursor. It supports local Whisper models, NVIDIA Parakeet, Apple's native Speech framework, and cloud providers (Groq, ElevenLabs, Deepgram, Mistral, Gemini, Soniox) — including real-time streaming for some of them. It optionally enhances transcriptions with LLM-powered AI before pasting. You can find all the information and download the official app from tryvoiceink.com.

What is Voiceitt?

Voiceitt is a speech recognition service that supports ß enables accurate real time voice transcription for people with speech disabilities (e.g., dysarthria, aphasia due to cerebral palsy, stroke, ALS, or other conditions), aging adults, and accented speakers. Users sign up and create a personal by profile with a minimum of 50 word training. They then can use the service in a web based app, Zoom or Microsoft Teams

What is this project & why does it exist?

Currently, this is a successful proof of concept that is evolving. It uses the source code of Voiceink and adds Voiceitt as a transcription provider. It is designed to be a drop-in replacement for the existing transcription providers, allowing for the use of the Voiceitt service across all applications on the MacOS platform with the addition of the post-processing available in VoiceInk. This work is never intended to be merged back into Voiceink , and the other transcription providers may be eventually stripped out so that this is only a client for Voiceitt.

This project developed out of personal passion and interest. I am a developer/technologist looking for work, who has cerebral palsy. I recently Googled out of frustration and found Voiceitt. I fell in love with it immediately -- it has been transformational. After discovering it had an API, I looked for ways to extend it to work in other situations. When I found Voiceink and the open source code for it. I started immediately trying to come up with a solution. Since I had been dabbling in AI (using Amp, I quickly found away develop this proof of concept.

Building for Local Use (No Apple Developer Certificate)

If you don't have an Apple Developer certificate, use make local:

git clone https://github.com/dmclark/VoiceInk
cd VoiceInk
sh scripts/sync-upstream.sh
make local
open ~/Downloads/VoiceInk.app

Note: The default branch is voiceitt so that upstream changes (VoiceInk) can continue to be merged in.

This builds VoiceInk with ad-hoc signing using a separate build configuration (LocalBuild.xcconfig) that requires no Apple Developer account.

How It Works

The make local command uses:

  • LocalBuild.xcconfig to override signing and entitlements settings
  • VoiceInk.local.entitlements (stripped-down, no CloudKit/keychain groups)
  • LOCAL_BUILD Swift compilation flag for conditional code paths

Your normal make all / make build commands are completely unaffected.

Requirements

  • macOS 14.4 or later
  • Xcode 15.3 or later installed
  • Voiceitt API key & APP_ID
  • Email & Password for Voiceitt user account (profile to use)

Configuration

Once you have the app running, configure Voiceitt to be the transcription provider.

  • Open the app and go to the settings tab.
  • Click on the "AI Model" section.
  • Click on the "Cloud" tab.
  • Select Voiceitt as the transcription provider. (Last option in the list )
  • Enter your Voiceitt API key and APP_ID.
  • Enter your Voiceitt email and password.

Documentation


About

Voice-to-text app for macOS to transcribe what you say to text almost instantly

Resources

License

Code of conduct

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • Swift 97.6%
  • Python 1.7%
  • Other 0.7%