Skip to main content
The layer generator turns a stack of trait layers into a collection, handling which combinations are allowed, how rare each trait is, and getting the result into your launch draft.

​
The flow

It reads top to bottom, in the order you’d actually do it:
1

Import your layers

A folder per layer. The order you put them in is the stacking order — the first one is the bottom of the picture.
2

Name your NFTs

Set the naming pattern. Anything you rename by hand keeps its name even if you re-run the pattern later.
3

Rules

Which trait combinations you don’t want. See below.
4

Collection attributes

Metadata that goes on every item.
5

Generate

Set how many you want and go.

​
Rules

A rule says: if these traits are present, these other ones can’t be. As you edit, you’ll see the effect immediately — how many combinations each rule removes, how many valid ones are left, and a filter over the preview grid to see what’s been excluded. Above about 30,000 combinations the running totals switch off, because working them out gets slow. The per-item “excluded by rules” marks still work.
Renaming a trait carries your rules with it. You don’t have to rebuild them.

​
Two different rarity controls

These get mixed up constantly, and they do different things.
The slider on a layer is the chance the whole layer appears. At 40%, roughly 40% of your collection has something from that layer and the rest has nothing from it. Items missing it get a “No ” mark in the grid.
Each image has a weight relative to the others in its layer. Weight 60 against weight 20 is three times as likely.You’ll see each one’s share as a percentage, a rough count, and a distribution bar. Equalize resets them all.Weight 0 turns the trait off entirely. It won’t appear at all — it’s shown dimmed with a marker so you can tell it apart from something merely rare.

​
Your artwork is built when you launch

The generator doesn’t make full-resolution images in your browser. It works out a recipe for each item and shows you a small preview; we build the real files during deployment. This does mean you can’t take finished artwork out of the tool without actually launching the collection. Both chains work this way — on EVM the finished image is pinned to IPFS during the deploy, and on Solana it’s built just before the Arweave upload.

​
How it picks combinations

Up to 100,000 possible combinations, every valid one is worked out exactly. So if your rules only leave 800 valid combinations and you asked for 1,000, you get told that precisely instead of a vague failure. Above that it samples. And if a very tight rule set exhausts the sampling budget, the last few get filled evenly from what’s left — so a very large collection against very restrictive rules can drift slightly from your intended weights at the tail.

​
The numbers

Collections generated before this system still work — they take the older path through launch.