MINTERA.TECH

How to Generate an NFT Collection From Trait Layers

August 3, 2026·7 min read

Turn a folder of trait layers into a finished NFT collection with matching metadata. How rarity percentages work, why uniqueness can fail, and what you get back.

What the generator does

The collection generator takes your trait layers and combines them into a finished collection. You upload one set of images per layer — backgrounds, bodies, eyes, headwear — say how common each trait should be, enter how many pieces you want, and download a folder containing every image and a matching metadata file for each one.

It runs entirely in your browser. Your artwork is never uploaded anywhere, including to us, which matters when the art has not been announced yet. The finished collection is written straight to your downloads folder.

Layers are drawn back to front

Each layer is one part of the artwork, and the order is the drawing order: the first layer is the background, and each one after it draws on top. Everything above the bottom layer needs a transparent background, or it will cover what is underneath. The generator checks this as you upload and warns you if an image has no transparency.

All images in a collection have to be the same size. That is also checked the moment you pick a file, rather than after you have uploaded two hundred of them.

Rarity percentages are exact, not approximate

If you set a trait to 5% of 1,000 pieces, you get exactly 50. Not roughly 50.

That is worth stating because it is not how every generator works. The common approach gives each piece a 5% chance of the trait, which across 1,000 pieces realistically lands anywhere between 36 and 64. You would then see 43 and reasonably conclude the tool was broken. Here the counts are decided up front and then dealt out, so the rarity table you see before generating is a guarantee rather than an estimate.

Percentages rarely divide evenly into your supply, so a few pieces get shifted to make the totals add up exactly. The generator tells you when it does that instead of adjusting quietly.

Why a collection can fail to generate, and what headroom means

Every piece has to be different from every other piece, and that is not always possible. Five layers of ten traits look like 100,000 combinations, but that number is misleading once your rarity settings are uneven. A layer where one trait takes 90% of the pieces is barely doing any work: nearly everything shares that trait, so it cannot tell pieces apart.

The panel under the supply field shows headroom, which is how many genuinely different combinations your settings reach compared to how many pieces you asked for. Under 2x you should expect a lot of near-identical art. Under 1x the collection cannot be built at all, and the generator says so before you start rather than failing partway through.

The fastest way to buy headroom is usually to even out the percentages, not to add more traits.

Rules stop combinations that look broken

Some pairs of traits should never appear together: a beard under a gas mask, earrings behind long hair, a hat clipping through antlers. Add a rule for each pair and the generator will never put them on the same piece.

Rules use up headroom, and they can make your rarity settings impossible to satisfy. If a trait needs 100 pieces but can only sit with 50 of them, you are told which two settings are fighting and what to change, before the run starts.

Hand-made 1/1s

You can drop finished pieces into the collection and give them specific token ids. The generator builds around them, so a piece you drew by hand can be #1 while everything else generates normally. Their traits are typed in by hand, because a 1/1 usually does not break down into the same layers as the rest.

One thing to know: rarity percentages apply to the generated pieces, not to your total supply. With 1,000 supply and 10 hand-made pieces, a trait set to 10% appears on 99 pieces, not 100. The rarity table you download reports both numbers so nobody has to guess which one you meant.

What you download

A zip containing:

  • images/ — one PNG per piece, named by token id.
  • metadata/ — one JSON per piece, in the standard ERC-721 shape that launchpads expect.
  • _metadata.json — every piece in a single array, for tools that want it that way.
  • rarity.csv — the trait counts and percentages, ready to share with your community.
  • config.json — every setting plus the seed. Re-import it with the same layer images and you get the same collection back, exactly.

Large collections arrive as several zip files rather than one, because browsers cannot reliably save a single file of that size. Unzip them all into the same folder; the numbering continues across them.

Taking it to launch

The folders are already in the shape the launchpad accepts, so you can upload them directly and deploy. Total supply has to match the number of pieces you generated.

You do not have to launch here. The download is a plain folder of images and standard metadata, and it works anywhere. If you do launch on Mintera, see the launch guide for mint phases and whitelists.

What it costs

Generating is free right now. If a fee is ever set, the price is read live from the contract and shown before you start, and it covers a run plus a number of re-runs so that adjusting your rarity settings does not cost you again.

Ready to dive in?

Explore live collections or launch your own on the Robinhood Chain.