This is a basic, simple prototype of the app Diane, which is a voice recorder using Whisper to transcribe and GPT to process text based on a template. Finally it saves the result to your Obsidian directory.
- Written in Python by a non-coder
- Only tested on MacOS
- Press "Spela in" to start recording your voice
- Press "Stoppa inspelning" - stops recording and saves the audio file to the audio directory and converts the wav file to mp3
- Choose a template from the dropdown menu which transcribes the audio file with Whisper and after that the transcribed text in sent to either GPT-3 och GPT-4 with a promt from the config.py file
- Press "Spara" (save) to save the result to your Obsidian https://obsidian.md/ directory
- You need an API key from OpenAI https://openai.com/
- Download files and directories to your computer
- Install FFMPEG on your computer
- Install python modules using pip : customtkinter, pyaudio, pydub and openai
- Open config.py and insert you OpenAI API key in OPEN_AI_API_KEY = '' and change the file path to your Obsidian directory, where you want to store your notes in OBSIDIAN_FILE_PATH = ''
- Run main.py and have fun
- Promtps for the templates are stored in config.py as well
- If you want to add or edit the dropdown menu it starts from line 272 in the function button_callback()
- If you run into empty recordings, you want to change the privacy settings of MacOS that prohibits the use the microphone via a python script that runs in IDLE. You'll also get a popup to grant microphone access for Terminal.
- version 0.1 | Initial Upload
- version 0.2 | Changed GUI fron tkinter to customtkinter. Audio file is converted to mp3 for file size which can be sent to Whisper (25 MB limit). Dropdown menu instead of buttons for templates.