How I built a music player with AI, from idea to publishing

One day I decided my site needed music. Not an embedded Spotify playlist, not a generic third-party widget. A player of my own, with my songs, my interface, my design. And I, who am not a software engineer, built it from idea to published with AI’s help.

Let me tell you exactly how it went.

Why a player of my own

WandaBuilds exists to prove that anyone can build things with AI, not just use apps others made, but create something of your own, with your rules. Music felt like the perfect test: it’s a real interface component, with state, with audio, with a database. There’s no more honest way to show this works than to show it working.

And since I built it without a programming background, I wanted that to be visible on the site.

The pieces of the system

Before writing a single line of code, I asked Claude to help me think through the architecture. That’s the first thing any builder does: understand what parts the system needs before building them.

The WandaBuilds player ended up like this:

• Supabase — the database where playlists and songs live, with title, artist, cover image, and audio reference.
• Cloudflare R2 — the storage where I keep the audio files. Think of it as Google Drive for code, but without the high costs of major cloud providers.
• Astro — the framework for the whole site, where the player component lives.
• Claude Code — the one that wrote most of the player’s code, the migration scripts, and the Supabase queries.

How I used AI in each part

The player component

I described to Claude what I wanted: a mini-player in the site’s footer, with play/pause controls, song navigation, and a progress bar. It asked me what framework I was using, how state worked on my site, what the visual style was. After that conversation, it gave me the component.

Did it work on the first try? Not entirely. I had to tell it what was breaking, what didn’t look right, how it interacted with the rest of the site. But that’s exactly what the process is: you iterate. It’s not instant magic, it’s fast collaboration.

The upload scripts

To upload songs to storage and register them in the database, I wrote some scripts with Claude. I explained the process in my own words: “I want to take an MP3 from this folder, upload it with this name, and save the title, artist, and public URL in the songs table”. Claude wrote the script. I tested it, told it what error came up, it fixed it. On the third try it worked.

The cover for each song

This was my favorite moment. Instead of searching Google Images, I generated the covers with Higgsfield: scene prompts that captured the mood of each playlist. Every song has its image. Every playlist has its album art. All on-brand, all consistent.

What surprised me about the process

The hardest part wasn’t writing code. It was describing precisely what I wanted.

When you tell AI “make me a music player”, the result is generic. When you say “I want a fixed mini-player in the footer that plays MP3s from a Cloudflare URL, with a next-song button, and the title shown on a single line with scroll if it’s too long”, the result is yours.

AI amplifies the clarity with which you think. Learning to describe what you want well is the most valuable skill of the modern builder.

What I learned from the process

You don’t have to understand all the code to build something that works. You have to understand what you want, know how to explain it clearly, and learn to read errors so you can tell the AI what went wrong.

That’s what being a builder today means: using the tools you have, iterating quickly, and shipping real things.

Can you build something like this?

Yes. Not the same player (that one already exists), but something of your own. A tool for your business, a component for your portfolio, a project you’ve been wanting to build and putting off because “you don’t know enough code.”

AI doesn’t ask you to know everything. Just to have the idea and the willingness to iterate until it works.

The player is already on WandaBuilds. Go to /music and listen. That’s what you can build.