Open source chord & beat detection application

Hi everyone,

I recently built ChordMini, an open-source tool that uses deep learning models and LLM to analyze songs and provide:

Chord recognition with 301 chord labels ( 12 keys x 25 types + N)
Guitar chord diagrams (currently no inversion labels)
Beat tracking and synchronized chord progression visualization (with metronome)
Lyrics integration (lrc & model transcription with music.ai api)
LLM used for further abstract analysis (key/tonal modulation correction, song analysis, and structural segmentation through color coded in the beat chord grid).

The webpage is at: ChordMini
The repo is at Github if you are interested.

Feedback, questions, suggestions are very welcome and any contribution is appreciated!

Comments

[Madza]

This is pretty impressive, familiar space to me! Congrats on the release! 🎉🎉

Been a music producer myself and built an audio player, radio, and music portfolio as well! 🎵

[Nghĩa Phan Trọng]

Thank you 🙂 . I'll check out some of your amazing work.

[Mikey Dorje]

My thoughts on this as a musician and as a developer. It is very impressive and very well made! I tested it out with a track with not-so-easy-to-detect complex chords, and it was surprisingly accurate. But the coolest thing is that with all the AI tools being created in the music field, this is actually very useful. A gigging musician who regularly has to learn new songs constantly would find this invaluable.

The majority of the music AI tools being created are doing musicians a disservice, not because they "replace jobs" (although that's part of it), but because they replace the creative process, which is itself the reward of music, not the final product created. More specifically, how the creative process utilises the brain and physiological functions and skills. For example, this tool does replace the skill of being able to learn a song by ear, which can be challenging but is also (because of its challenging aspects) a really remarkable and self-rewarding skill and experience. But again, a busy, gigging musician would find this very helpful—they would likely already use apps with huge catalogs of popular songs or traditional chord books for this anyway.

All in all, very cool and easily one of the most actually useful AI music tools I've seen! I plan on checking out the GitHub repo! Really interested in seeing the inner workings of this.