Pick a chord or a scale and watch it appear on the fretboard β then tap any note to hear it. When you're ready, switch to Ear Training and start learning to hear what you play. No downloads, no audio files. Every sound is made live in your browser.
Standard tuning, low to high: E A D G B e. The board runs from the open strings to the 15th fret. If you're working on a shape, try this: pick the chord, hit Strum, then tap the notes one string at a time from low to high and say the note names out loud. Boring, works every time.
You'll hear one note. Tap where that note lives on the fretboard below β any position with the right note name counts.
Hit play, listen hard, then tap your answer on the fretboard. You can replay as many times as you need β no penalty.
Tap a note on the board to answer. Every position is playable, lit or not.
Five minutes a day beats an hour once a week. Start with Guess the Note until you're hitting 7 or 8 out of 10, then mix in chords. The trick that helped me: before you tap, sing the note back and try to picture where your fingers already know it from a song you play. Your ear usually knows more than you think β it's the connecting it to the fretboard that's the work.
Hi, I'm Rick. I play guitar β American Strat, mostly classic rock and metal, Frusciante is my guy β and I've been meaning to get serious about two things for a while: actually knowing the fretboard instead of just knowing shapes, and training my ear. I also wanted something public I could point people to that shows how I work with AI to build software. Not a tutorial I copied. A real thing I wanted, built by talking to an AI until it existed.
So this is that. One page, no sign-up, no downloads. Pick a chord or a scale and it lights up on the fretboard and plays. Then there's an ear training game that plays you a note or a chord and makes you find it. This page is the live demo, and everything below is honestly how it got made.
I didn't start with a spec. I started with a question. I asked the AI to brainstorm with me:
"I want to have something public on my GitHub that shows I know how to use AI for coding, vibe coding, etc. β without exposing my other sites. What's a cool simple thing to make via vibe coding that's simple to understand? Anything with guitar, comics, video gamesβ¦ something you can spin up and showcase and then the person can click through?"
It came back with three ideas: a "what should I play tonight" game picker, a comic pull-list tracker for my Fantastic Four / Spider-Man / Wolverine runs, and this fretboard thing. The fretboard won because it was the only one I'd actually use every week. That's my first rule with this stuff now: build the thing you'll open again on Tuesday, not the thing that just demos well.
My instinct is to ask for everything at once. I've learned that gets you a mess. So the next prompt was about cutting it down:
"Okay, the fretboard explorer. For version one I just want: pick any chord or scale, see it light up on the fretboard, and hear each note. And an ear training mode β it plays a note or a chord, I tap my answer on the fretboard, and it keeps my score and streak. Two modes to start: guess the note and guess the chord. Keep it to one page."
The AI pushed back in a useful way here, which is something I look for now. It pointed out that "tap your answer on the fretboard" works great for a single note, but a chord has a root, a quality (major, minor, 7thβ¦), and about five places to play it β so for chords it suggested tapping the answer as a choice between four chords, then lighting up the shape on the board so you still learn where it lives. I went with that. Good AI collaboration isn't the AI saying yes to everything. It's it catching the awkward bit before you've built it.
The thing I assumed would be the hard part was sound. I figured we'd need guitar samples, and then licensing, and file sizes, and loading spinners. I asked:
"For the sound, do I need to find guitar recordings somewhere? I don't want to deal with audio files if I can avoid it."
Turns out no. Browsers can make sound from nothing using something called the Web Audio API. Every note on this page is generated live with a bit of code β an oscillator for the pitch, a filter and a volume envelope shaped so it decays like a plucked string instead of beeping like a microwave. Is it my Strat through a good amp? No. Is it clearly a note, in tune, the instant you tap, with zero files to load? Yes. For ear training that's honestly the part that matters. This was the biggest lesson of the whole build: ask the "dumb" question early. The thing I thought would take a weekend took about twenty lines.
This is the part I actually wanted to understand instead of just accepting, because if I couldn't explain it, it wasn't really my project. So I made the AI walk me through it, and it's simpler than it looks:
open string number + fret. Third fret on the A string is 45 + 3 = 48, which is C.That's the whole trick. Chords, scales, the quiz answers β all of it is that one idea with different recipes. Once I got that, I could start asking for changes in a smarter way ("add a sus4, it's root + 5 + 7") instead of just saying "make it better."
It was not one magic prompt. It never is. Roughly, the loop went like this:
| Piece | What it is |
|---|---|
| The page | One single HTML file. The layout, the styling, and all the logic live in it. No framework, no build step, nothing to install. I can open it, read it, and change it β that was deliberate, it's how I like to work. |
| The sound | The Web Audio API, built into every browser. Notes are synthesized live β no audio files anywhere in the project. |
| The music theory | About ten chord recipes and eight scale recipes, stored as lists of numbers (intervals from the root). Adding a new chord is adding one line. |
| Hosting | Free static hosting (GitHub Pages). Push the file, it's live. Public repo, so anyone can read the code and this README. |
| Score & streak | Saved on this device, so your score and best streak are still there when you come back. No account, nothing leaves your browser. |
I didn't write most of the code in this project, and I'm not going to pretend I did β that would defeat the entire point of putting it on my GitHub. What I did was decide what to build, cut it down to something finishable, learn the one idea the whole thing runs on, give specific feedback when it wasn't right, and check the output like it mattered, because a teaching tool with wrong notes is worse than no tool.
That's what "vibe coding" actually looks like when it works, at least for me. The AI is fast at typing. I'm the one who has to know what good looks like. And now I have a fretboard in my pocket that I use to practice my ear, which was the point all along.
β Rick