Inspiration
DJs, raves, etc. We want to be DJs but creating our own backgrounds is tedious.
What it does
DJ Background Generator creates music videos with time-synced lyrics and AI-generated backgrounds. It takes a song file and word-timed lyrics in LRC format as input, then produces a complete video using Blender where each lyric appears with a unique AI-generated surreal background synchronized to the music.
How we built it
There's two components to this project: the Blender-based video generation and rendering and the custom ComfyUI/ControlNet based background image render.
The GitHub repo has a pipeline that
- Parses LRC subtitle files to extract time-synced lyrics
- Generates a Blender project with text objects for each lyric
- Renders individual frames for each lyric line
- Submits rendered frames to a custom pipeline in ComfyUI for AI background generation
- Returns the frames and renders them into a music-synced background video with lyrics
Challenges we ran into
- Achieving precise time synchronization between lyrics, audio, and visual transitions
- ComfyUI is confusing, slow, and pretty expensive
- AI's don't do a good job interacting with Blender
Accomplishments that we're proud of
Lots of scripts doing almost fully automated interaction with Blender!
Custom ControlNet + image pipeline, as a service!
What we learned
We gained experience integrating Blender's Python API for programmatic 3D rendering, working with ControlNet and image models, and figuring out how to subtitle and time-sync lyrics really fast.
Log in or sign up for Devpost to join the conversation.