AI AdventureAICreate with AIDay 8
🌫️ Create with AI · Day 8

Smash the Picture

Weird secret of AI image makers: before they can build a picture, they learn to destroy one. You add a tiny bit of TV-static noise, then a bit more, then more… until the picture melts into pure random fuzz. Grab the slider and dissolve a picture with your own hand.

↓ melt it
🌫️ Lab · The Noise Slider

Drag it. Watch it dissolve.

Pick a picture, then drag the noise slider from 0% to 100%. At every step we mix in a little more random static. Same picture, slowly turning to fuzz — live, in your browser, one pixel at a time.

the real picture (0% noise)
noised: 0%
less noise0%pure static

Every notch replaces a bit more of the real picture with gaussian static — the same kind of random fuzz used in real AI image models. Slide slow and watch the picture fight to survive. 🌫️

▶️ Play · Step by Step

Add noise one step at a time

Real models don't dump all the noise at once — they add a little at each timestep, following a plan called the noise schedule. Press play and watch 12 steps go by, from crisp picture to total static. Numbers show exactly how much noise is in the picture at each step.

step 0 / 12 · 0% noise
Notice the pace: the schedule adds noise gently at first, then faster. That smooth ramp is a real design choice — it makes the reverse trip (building a picture back, Day 9!) way easier. 📈
🐞 Break-It · Everything Melts the Same

At 100%, every picture is the same nothing

Here are three totally different pictures, all cranked to 100% noise. Press the button to re-roll the static. Can you tell which was the sunset? The smiley? You can't — and that's the whole point.

was: 🌅 sunset
was: 🙂 smiley
was: 🏔️ mountains
The big idea: every image melts into the same meaningless static. That sameness is the secret weapon. To make a brand-new picture, a model just starts from this random static and runs the whole melting process backwards. Noise → picture. That's Day 9. 🔮
🏷️ The pro wordsSlowly wrecking a picture with static is called forward diffusion. The static is gaussian noise (bell-curve randomness). The plan for how much to add at each step is the noise schedule, and each little step is a timestep. Grown-ups say a picture gets "noised over T timesteps" — you just did it by hand.
🎯 Boss · Name That Noise Level

How much noise is hiding in here?

We noised this picture by a secret amount. Study it, then drag your guess slider to how much static you think is mixed in. Get within 15% of the truth and you score. Nail it 3 times to win the badge!

🤔 secret noise level…
0% (clean)50%100% (static)
Correct: 0 / 3
🚀 Beyond the Basics

The dumbest part is the whole plan

Here's the twist that trips people up: the forward process — the melting you just did — involves zero learning. It's fixed math. Add gaussian noise, follow the schedule, done. A calculator could do it.

🌫️

Forward (today)

Picture → static. Fixed recipe, no brain needed. This is just so the model has millions of "noise → clean" examples to study.

Reverse (Day 9)

Static → picture. THIS is where a huge neural net trains for weeks to guess "what noise do I remove next?" The actual magic.

🎨

Steered (Day 10)

Same reverse trip, but a text prompt whispers "make it a dragon" at every step. That's text-to-image.

🏷️ Why no giant model here?Real image models need a big GPU and gigabytes of weights. But the forward math is tiny — so everything on this page is the exact same noising math real studios use, just running on a 220×220 picture so it flies on any laptop. No download, no cheating. 💪

🎉 Day 8 Complete!

You learned forward diffusion: how models wreck pictures on purpose, one gaussian-noise step at a time, until every image becomes the same random static. That "everything melts to the same nothing" fact is the launchpad for making brand-new art.

Day 9: Un-Smash the Picture →
Tomorrow: we run the whole thing backwards. Start from pure static and watch a picture rise out of the fuzz, step by step. That's how AI actually paints. ✨
AI Adventure · Create with AI · Day 8 — made for young inventors 🚀