This repo creates a little app that lets me have a podcast with ChatGPT as my co-host. The codebase is awful but functional as is everything I make.
- Clone the repo to your local machine
- Good luck figuring out all the required python packages
- Open
config.py
and paste in your OpenAIAPI_KEY
. Also feel free to edit the prompt or model being used.
- Navigate to the repo in your terminal
- Run
python main.py
- Hit the
Start/Stop Recording
button to begin recording. If your mic has two channels this whole thing will probably break. - Hit the
Start/Stop Recording
button to end recording. - Hit the
Transcribe Recording
button to send the recording through OpenAI's WhisperAPI to transcribe it. - Hit the
Get Response
button to send the transcription through ChatGPT for a response. - Hit the
Create & Speak Response
button to create an audio recording of ChatGPT's response and have it spoken out loud to you.- I am sorry for the shitty tts engine I used, planning to switch to a high quality API that used a transformer model in the future.
- Repeat steps 3-8 until your podcast is over
- Share the files
full_recording.wav
andfull_transcript.txt
on your podcast platform of choice.