By Matt Shumer (https://twitter.com/mattshumer_)
Generate a fictional podcast in minutes with AI.
This is not for commercial use -- purely for research and fun!
-
Dynamic Participants: Define the participants in your podcast. Be creative!
-
Voice Cloning: The system clones the voices of the participants from provided YouTube clips. This makes your podcast sound like an actual conversation between the participants.
-
Open the notebook in Google Colab or in a local Jupyter notebook.
-
Define your podcast participants, topic, and number of turns. For example:
speaker_one = "Joe Rogan"
speaker_two = "Rick Sanchez, of Rick and Morty"
topic = "Artificial Intelligence"
number_of_turns = 7
- Provide YouTube clips and voice descriptions for the participants:
speaker_one_clip = "https://www.youtube.com/watch?v=CM_LWxh33Z8" # a video of speaker_one speaking
speaker_one_voice_description = "American male, deep voice."
speaker_two_clip = "https://www.youtube.com/watch?v=GKPcHInn14c" # a video of speaker_two speaking
speaker_two_voice_description = "A bit gruff and raspy, often slurred due to his constant state of inebriation. His speech is punctuated with frequent burps and stammers. He speaks with a cynical and sarcastic tone, often sounding dismissive or condescending. His voice also has a certain manic energy to it, reflecting his chaotic personality."
- Run the cells to generate the podcast.
This project is MIT licensed.
Matt Shumer - @mattshumer_
Project Link: https://github.com/mshumer/gpt-podcast
Lastly, if you want to try something even cooler than this, sign up for Personal Assistant (most of my time is spent on this). It's basically an AI that can operate your web browser to complete tasks for you.