Check out my project on GitHub: https://github.com/pncnmnp/phoenix10.1
And here's a demo on SoundCloud
For the past few months, I've been hard at work on a project to create personalized radio stations with a text-to-speech based radio jockey, and I'm excited to share that it's finally in a polished state!
So, what does it do? Well, it's a radio station that plays your songs (even from your favorite artist/genre), finds and automatically plays interesting clips from preferred podcasts, provides weather updates, delivers daily news, and even brightens your day with fictional company ads, and much more.
Now, how did I make this magic happen? Here's the breakdown:
- The text-to-speech part is powered by Coqui-ai's library.
- Prompts spoken by the radio jockey are generated using ChatGPT and character.ai with manual selection for that personal touch.
- Songs are generated using
ytmdl
and the metadata is vetted from libraries likebillboard.py
,musicbrainzngs
, anditunespy
. - To find interesting clips from podcasts, it looks for clips separated by significant silence and selects the ones with the maximum running time, based on a threshold.
- Weather and news are obtained using wttr.in and RSS feeds respectively.
- Plus, it generates fascinating cover art using randimage, adding that extra visual flair!
And voila! After lots of code optimization and testing, you've got yourself a fully automated radio station.
I'd love to hear your thoughts on it!
Note: The awesome cover pic is from Mehmet Turgut Kirkgoz.
Top comments (0)