Hello world

• A free, cross-platform app that automatically write what you speak to current opening window
• Context: I was learning C and tired of having to write notes all the time, so I decided to create this app.
  I also saw some similar apps, but they are not free, so I decided to make a free version for myself and also test the new whisper model by OpenAI :)
• I hope this app will make me a better English speaker as speaking skill is my weakness :)
• I'm using both Windows and Ubuntu so Windows will be the first targeted platform, it will be easy to port to other OS (in my dream)
• For future release, I want to make a C/C++ version of this app for learning how to program in C/C++ :)

Installation

• Requirements: Python 3.10+, Windows 10, atleast 8GB of RAM
• You need to install python libraries first: pip install -r requirements.txt
• Optional: jupyter notebook: pip install notebook

Windows

• Install chocolatey: https://chocolatey.org/install
• Install ffmpeg: run Power Shell with administrator: choco install ffmpeg
• Install Rust: pip install setuptools-rust
• You are good to go :)

Notes

• The default model is "small", we may need to change it later if the voice detection's accuracy is bad
• I have no knowledge about audio/speech processing, so helps/advice from other folks are welcomes

Todos

• Delete audio files after transcribe
• Remove hardcoded values
• Better text aligment
• Add support for Ubuntu, MacOS, Iphone, Android, Windows Phone, Rasberry Pi, Andruino
• Try other apps and compare the performance/accuracy
• Test the voice detection accuracies (on-going)
• Auto use GPU if possible (on-going)
• Auto detect RAM and choose the most optimal model size
• Test model with math symbols
• Listen for sound and transcribe concurrently
• Adaptive voice threshold