Lo-FAI

• 

  A screenshot of the livestream in progress

Inspiration

We were inspired by 24/7 lo-fi music streams and wanted to create something that allowed for more user interaction

What it does

Lo-FAI is a 24/7 stream that generates AI music based on user input through Twitch chat.

How we built it

• GitHub
• OBS (to live-stream on Twitch)
• Python
• Twitch API (for the chatbot)
• Riffusion (for the music)
• Stable Diffusion (for the images )

Challenges we ran into

• Integrating the three main parts of the project (Visuals, Music, and the Twitch chatbot)
• Minimizing delay and playback and getting it to run in real-time
  
• Getting the Twitch API to interact with our main script
• Getting Stable Diffusion and Riffusion to run together (running two instances of Stable Diffusion)

Accomplishments that we're proud of

• Images are very pretty
• Managed to persevere through 15 hours of integration
• Seamless transition between genres
• Upscaler greatly improved the audio quality

What we learned

• Integrating complex parts of a program at the last minute is hard
• One of our team members (who will not be named) learned not to publish API keys to a public github repository
• Delegating work among team members allows for more work to be accomplished, especially if we help each other and work together

What's next for Lo-FAI

• Generating "album cover" art based on the user submitted prompts
• Allowing for more time between prompt changes
• (If partnered) change from a Twitch chat bot to a chat poll

Built With

• github
• obs
• python
• riffusion
• stable-diffusion
• twitch