This project will generate one episode a podcast called "No Humans Were Involved with this Podcast" when run. The structure of each episode will be as follows:
Each segment is generated using a separate prompt. The prompts for this system are (currently) human generated, but all of the episode content is written and recorded by a generative AI. The episode currently run 20-30 minutes in length and are generated as .wav files. Additionaly, the script for each episode is recorded as a .txt file.
All of the generative models being used in this project are provided by Hugging Face. The list of models used is:
This project is under heavy development and you WILL experience difficulty running it yourself. You may find yourself having to build an appropriate directory tree and tweaking a line or two.
Having said that:
python3 main.py
...is all you need to run. The main file will run the others as necesary.
Anecdotally, this program takes about 12 hours to run on an 8-core CPU with an RTX 4070. Your milage may vary.
I have added some sample output mp3s to this directory. Since development on this project is ongoing, these samples are not necesarily good examples of the current output of the systemn.