Chord-O-Matic: Upload a Song, Learn Its Chords
You want to play along with a song. If it is a hit, somebody has already worked out the chords and posted them. For everything else you work it out by ear.
Chord-O-Matic takes the MP3 instead. It listens, writes down every chord it hears, and puts each one above the lyric word it lands on. Plug in a MIDI keyboard and it will run you through the song and keep score.
Designed by me, built by directing AI: the Python and React are AI-written under my direction. I planned it, and I reviewed and tested it.
Why it exists
I wanted to hand the computer the song itself and get the sheet back.
Then it kept going. Once the app knows every chord and when each one lands, it can do more than print a sheet. It can sit at the keyboard with you and teach the song one chord at a time. That is where Practice Mode came from.
What it does
Upload a song, get the sheet
Drop in an MP3. The app reads the title and artist from the file’s own tags, finds every chord change with its timestamp, and shows its progress as it works.
Chords above the right words
The classic chords-over-lyrics layout, held in line by a monospace grid. Synced lyrics give each line its timing, and each chord lands on the word playing when it changes. If the lookup comes back empty, paste the lyrics in yourself.
Diagrams for your hands
Small fingering diagrams sit above the chord symbols, guitar or piano, your pick. When chords crowd together the diagrams stagger into rows instead of overlapping.
Transpose on the spot
Shift the whole song up or down a semitone at a time. Every chord symbol and every diagram follows.
Practice Mode
Plug in a MIDI keyboard and the app steps through the song chord by chord, listens, and keeps score. Follow along, or flip to call and response: it plays the chord, you play it back. A strict setting demands the exact notes and nothing extra.
A library, day or night
Every analyzed song is saved to a library of album-art tiles behind a simple sign-in. A night mode turns the app dark, and it remembers your choice.
How it works
I directed AI to build the chord detection. The audio is boiled down to a moment-by-moment fingerprint of which of the twelve notes are sounding, scored against a template for every major and minor chord. A smoothing pass votes away the jitter, and the winners merge into timed chord events.
Lyrics come from an open synced-lyrics source with a timestamp on every line. Each chord attaches to the word whose slice of the line contains its start time, which puts the chord above the right word instead of the right verse. Practice Mode closes the loop: the notes you press arrive over MIDI, get reduced to bare note names, and are checked against the target chord as you play.
I settled for two things on purpose. Detection gets roughly 70 to 75 percent of the chords right, which was good enough for a first version. And lyric timing is per line, not per word, so a chord’s placement is close, not exact.
Where it is headed
I set out to print a sheet. The keyboard trainer it turned into is the part I am proud of.