Blob Opera - Playful AI-powered opera you control with your mouse

What Is Blob Opera

So let me guess — you love music, but you're convinced you don't have the "talent" for it. You never learned an instrument. You can't read sheet music. And singing? Let's not even go there. The idea of actually making music feels like a door that's permanently locked for people like us.

But here's the thing — I found this beautiful little loophole.

It's called Blob Opera, and it's a free, web-based machine-learning experiment that turns music-making into something ridiculously fun. No microphone. No notation. No "correct" way to do it. You just move a bunch of colorful, squishy blobs around your screen, and they sing for you. That's it.

Created by artist David Li in collaboration with Google Arts & Culture and released in late 2020, Blob Opera is built on real machine-learning technology. Google trained the system using 16 hours of recordings from four professional opera singers — tenor Christian Joel, bass Frederick Tong, mezzo-soprano Joanna Gamble, and soprano Olivia Doutney (with additional help from Ingunn Gyda Hrafnkelsdottir and John Holland-Avery).

Here's the cool part though: you never actually hear their recorded voices. The AI doesn't just play back clips of these singers. Instead, it was trained on their performances to generate its own interpretation of the four voice types. The blobs produce brand-new sound every time you move them, filtered through everything those singers taught the model about opera.

Now, one quick clarification — there's the official Google-hosted experiment and then there's an independent fan site at blobopera.xyz. Google hosts the actual game you play. The fan site adds guidance, tutorials, and learning tools around it — think of it as the helpful friend who walks you through the experience.

Best part? It's completely free, no account needed, and runs right in your browser.

Blob Opera's Core Features

You're the conductor, the blobs are the choir

The magic of Blob Opera is how simple the controls are. Each blob represents one voice — a bass, a tenor, a mezzo-soprano, or a soprano. Move a blob up and down and the pitch changes: higher for a higher note, lower for a deeper one. Move it left and right and the vowel quality shifts, changing the tone without you ever needing to know what a phonetic symbol is.

That's the whole control scheme. Two dimensions, infinite possibilities.

Instant four-part harmony

Here's where the machine learning gets wild. The second ML model handles real-time harmony — the moment you take control of one blob, the other three automatically respond, harmonizing with whatever you're doing. You don't need to compose a chord progression or arrange four voices manually. You just move one blob, and boom — you've got a full four-part vocal texture happening in real time.

A playful intro to opera voices

The four blobs aren't random colors — they're actually a proper operatic quartet. Bass sits at the lowest register, tenor above that, mezzo-soprano in the lower female range, and soprano on top. It's a genuinely fun way to start hearing the difference between voice types, even if you've never thought about "registers" in your life. Think of it as a playful introduction to voice classification, not a professional vocal lesson.

Record, share, relive

One of my favorite features: you can record your performance. The recording preserves your actual gestures, so when you share a link, anyone can replay your exact performance from the beginning. No account needed, no sign-up friction. It turns your messy, experimental play into little structured compositions — press record, shape something with a beginning, a middle, and an end, then share it with a friend who'll probably say "how did you make this?"

Preset songs and a world tour

Blob Opera started its life as a holiday experiment with optional seasonal songs. It later expanded into a world-tour version, adding backgrounds and local songs from cities like Cape Town, London, Mexico City, New York, Paris, and Seoul — you'll hear everything from "Frère Jacques" to "La Bamba." The presets show off what the voices can do, but honestly, free play is where the real magic lives.

It's about exploration, not performance

There's no fixed score, no wrong answers, no "finishing" the game. The whole point is experimentation — moving slowly, pausing to listen, hearing what happens when the harmony shifts under your control. You're not performing. You're exploring.

One honest note on tech requirements: Blob Opera relies on modern web audio technology, so it may lag on older devices. If that happens, close other audio tabs, make sure your browser is up to date, and try headphones to separate the individual voice parts.

Who Is It For

So who's this actually for? Let me paint a few pictures.

You, the music-curious beginner. If you've always suspected there was a musician inside you but never had the tools to find out, this is your gateway. In five minutes you'll be hands-on with register, pitch, and harmony — concepts that normally take months of music theory to grasp. You'll feel them before you ever need to name them.

Parents and teachers with kids to entertain. Blob Opera is a guilt-free free browser tool that makes musical ideas genuinely audible. A kid can see that moving the blob up makes the note higher, that changing position changes the vowel sound, that all four voices can move together or break apart. It's a toy that happens to be educational, which is the best kind of toy.

The music-lover who just wants to unwind. After a long day, there's something deeply relaxing about dragging a blob through space and watching a full choir respond. It's a creative sandbox with zero pressure. No score to beat, no level to clear.

The tech and ML nerd. If you're curious about how machine learning actually works in the wild, here's a real convolutional neural network trained on 16 hours of singing, plus a second model doing real-time harmony — all wrapped in something you play with for fun. It's AI demystified, one blob at a time.

Museums and education programs. As part of the Google Arts & Culture family, Blob Opera fits naturally into cultural and educational settings. It's approachable, interactive, and culturally playful — a great gateway to opera for people who've never stepped near one.

But let me be clear about what it's not. Blob Opera is not a voice simulator, it's not a vocal training tool, and it's not a scientific model of real singers. It's an interactive, opera-inspired experiment. If you want to learn how to sing like an opera star, this isn't that. But if you want to play with the idea of opera and hear what harmony sounds like, it's perfect.

How to Use Blob Opera

Okay, let's get you playing. Here's the whole process, from zero to beautiful noise in about a minute.

Step one: Open blobopera.xyz (the fan site) or go straight to the official Google experiment. Hit play.

Step two: Pick a blob. Any blob. They're all fun. I'd suggest starting with the bass or tenor because the lower registers are easier to hear at first, but honestly, just grab whichever color catches your eye.

Step three: Drag it around. Up and down changes pitch — the higher you go, the higher the note. Left and right changes the vowel sound, giving you subtle shifts in tone quality. Hold a position steady for a moment and you'll hear the full tone color of that specific voice at that specific point.

That's it. You're making music now.

Want a five-minute musical journey? Here's a tiny loop to try. First, find the lowest possible note and the highest possible note, just to feel the range. Then pick a spot and hold it still — listen to how the other blobs respond harmonically. Now make a slow change, sliding the pitch gently upward, and notice what happens to the texture of the whole ensemble. The blobs aren't just following you — they're reacting, creating tension, stability, and balance shifts that they choose.

Try the presets. Hit a preset song button and watch how each voice behaves. You'll hear what the quartet is capable of, and it gives you ideas for your own experiments.

Record something small. Press record, keep it to about ten or fifteen seconds, stop, and listen back. Then change one thing and try again. That's the core of composing — listen, revise, repeat. Short attempts, one clear change per revision, and you'll be amazed how quickly you build something you're proud to share.

If it sounds broken or silent, here's your checklist: close any other tabs playing audio, switch to a current browser version, make sure you're on the official Google experiment page, and try wearing headphones so you can actually separate the four voice parts.

The Tech Behind the Blobs

Okay, let me pull back the curtain, because the machine learning behind Blob Opera is genuinely fascinating — and you don't need a CS degree to get it.

There are two models working together:

Model one is a convolutional neural network trained on 16 hours of real opera singing. Its job is to reproduce each of the four voice types — bass, tenor, mezzo-soprano, and soprano. It learned from the singers' expressive examples and musical knowledge, essentially absorbing how they sing rather than just what they sang.

Model two handles real-time harmony. When you take control of one blob, this model instantly makes the other three respond harmonically. You're not explicitly telling them what to do — you're just moving one voice, and the AI figures out how the other three should fit around it. In real time. In your browser.

Here's the key insight that a lot of people miss: the sounds you hear are AI-generated, not recorded clips. The quartet's voices shaped the model's understanding of opera, but every note the blobs produce is the AI's own interpretation inspired by those recordings. That's why the same gesture can sound different every time — the model is generating, not replaying.

This design choice has some brilliant upsides and a few honest limitations worth knowing about.

That trade-off — playful generation over faithful reproduction — is exactly what makes Blob Opera so approachable. It's not trying to be a realistic simulation. It's trying to make you feel like you're conducting an opera, and it succeeds by turning machine learning into something you can touch and hear.

Common Questions