How Many Traits Do You Need for 10,000 NFTs? The Real Maths

The short answer

Multiply the choices in each folder. The result is the most different NFTs your traits can ever make, and it has to be bigger than your collection size, usually by a wide margin. Three things make the real number smaller: never-together rules, traits that hide other traits, and rare traits that need room to spread out.

Start with the multiplication

Every NFT takes one trait from each folder, so the choices multiply. Here is a typical six-folder character collection:

691,200 is 69 times 10,000, so this set has lots of room. Four folders of 10 traits, on the other hand, make exactly 10,000, which leaves no room at all (see “How much room is enough?” below).

What a None choice adds

A “None” choice lets a folder sometimes be empty, and it counts as one more choice. Add None to the 12 hats and that folder has 13 choices: 8 × 6 × 15 × 10 × 8 × 13 = 748,800.

An NFT that gets None simply has no hat, and its metadata leaves the Hat trait out.

How a never-together rule removes combinations

A never-together rule stops two traits from appearing on the same NFT. Say one hat must never meet one piece of clothing. Every combination holding both is gone: all the backgrounds, bodies, eyes and mouths that could have gone with that pair, 8 × 6 × 10 × 8 = 3,840 combinations. 691,200 − 3,840 = 687,360.

One rule barely matters in a set this size. Hundreds of rules on a small set can matter a lot, which is why the studio counts again every time you change one.

Traits that hide other traits

Multiplying counts lists of traits. What collectors see are pictures, and two different lists can make the same picture.

A trait that fills the whole picture

Say one piece of clothing is a full costume drawn over the whole canvas, covering the background and the body. NFTs wearing it can only differ in the eyes, mouth and hat drawn on top: 10 × 8 × 12 = 960 different pictures, even though they hold 8 × 6 × 10 × 8 × 12 = 46,080 different lists of traits.

So the real total is 645,120 NFTs without the costume plus 960 with it: 646,080, not 691,200. Puffles Studio counts it this way. A trait image with no see-through pixels, in any layer above the bottom one, is treated as hiding everything under it. That is also why every layer except the bottom one should have a transparent background (how to prepare your layers).

A trait that hides part of another

A mask over the eyes can cover the pupils. Two eye traits that differ only in their pupils then look the same under that mask, and two NFTs that differ only in those eyes are the same picture. The studio finds these pairs itself: it compares the images on a 128 × 128 grid wherever the upper trait is solid, and counts each such pair as one.

Why a list of traits is not a picture

Code generators such as the HashLips Art Engine check uniqueness with a “DNA” made from the traits each NFT gets. That catches repeated lists, not repeated pictures. Puffles Studio checks pictures: while planning, it knows which traits hide others, and after building it compares every finished image and warns you if any two came out identical.

Rare traits need room

Room is not only about the total. A common trait also needs enough different partners in the other folders.

Take five folders of 4 traits: 4 × 4 × 4 × 4 × 4 = 1,024 possible NFTs. Asking for 1,000 with even weights works. Now give one trait a weight of 10 while the other three in its folder stay at 1. It would have to appear 769 times, but only 4 × 4 × 4 × 4 = 256 different NFTs can carry it. The studio refuses and says so:

“A / A 1” would appear 769 times, but the other folders only allow 256 different NFTs with it. Lower it, add traits to the other folders.

It names the trait instead of quietly making duplicates.

Weights that round to zero

The opposite problem is a trait so rare that, at your collection size, it rounds to no copies at all. In a test at 1,000 NFTs, a hat with a weight of 0.0001 next to eleven hats at 1 got 0 copies, and the studio warned:

“Hat / Hat 1” is so rare at this collection size that it rounds to 0 and will not appear. Give it an exact count to force it in.

Switched to an exact count of 7, it appeared exactly 7 times. NFT trait rarity for creators explains how weights and exact counts work together.

How much room is enough?

There is no fixed ratio. Here is what the studio’s own planner did in our tests on 22 September 2026:

• With 691,200 possible and 10,000 wanted, the plan was ready in well under a second, with every weight left as it was.
• Four folders of 10 traits make exactly 10,000. Asking for all 10,000 with even weights worked, but only because every trait was used exactly 1,000 times. Giving a single trait a weight of 2 was refused: it would need 1,819 copies, and only 1,000 different NFTs could carry it.
• In the same set, 9,975 NFTs from traits plus 25 1 of 1s did not finish. The studio stopped with “Could not finish: 2 NFTs are still duplicates” and suggested evening out the weights, adding traits or lowering the size.

The planner searches for an arrangement; it doesn’t prove one exists. Right at the limit it can stop without an answer. Aim for many times your collection size, and more if you plan rare traits or lots of rules.

Check your own numbers

In Puffles Studio the number of possible unique mixes appears as soon as you import your folders, and it changes as you add None choices, rules and 1 of 1s. It counts pictures, as described above. 1 of 1s count toward your collection size, so 10,000 with 25 1 of 1s means 9,975 made from traits. The largest collection the studio makes is 20,000, for free (how other generators price it).

The studio needs a free sign-in with an email or a wallet. How to make a 10,000 NFT collection walks through the whole process, and the guide’s preview step shows where the count lives.