Gustav Most NFT collections that launched during the boom were assembled, not designed. A set of base characters, a stack of randomised trait layers, a script to generate thousands of combinations, and an OpenSea listing. The images were functional but forgettable - and when the market stopped rewarding novelty, forgettable was exactly what they were.
My own collection took a different starting point. Three characters: Alien, Human, Robot. Each one rendered individually, with distinct visual identity, deliberate lighting, and character design that would hold up whether you were looking at a 24x24 thumbnail or a high-resolution print. The goal was a body of 3D work I would be proud to have under my name, independent of what the market was doing.
Starting with archetype, not attribute
The first decision was conceptual: what is this collection actually about? A lot of NFT projects start with "what traits can we generate?" rather than "what is the core idea?" The generative approach produces variety but it rarely produces meaning.
I started with three archetypes that between them cover a version of the whole human story: the alien (the unknown, the other, the extraterrestrial), the human (the familiar, the self, the baseline), the robot (the made thing, the successor, the question about what comes next). Each character is a distinct design problem with its own visual language - which is a much more interesting brief than "helmet: 12 variants, background colour: 18 variants."
The 3D production process
Each character was modelled and textured from scratch. The card format - the NFT card aesthetic that gives each piece a framed, collectible quality - places specific demands on the composition. The character has to read at small sizes. The background has to complement the subject without competing with it. The colour palette needs to be consistent enough across cards that they clearly belong to the same collection, while being distinct enough that each card has its own identity.
Lighting was probably where I spent the most time. A card format is essentially a portrait, and portrait lighting in 3D follows most of the same logic as in photography: key light, fill, rim, and the relationship between them determines whether the subject feels alive or flat. With a robot character, you have additional decisions about how the materials interact with light - specular reflections, the sharpness of highlights on different surface types, how emissive elements (screens, LEDs) integrate with the lighting rig without blowing out the render.
The best NFT art is indistinguishable from the best digital art in any other context. If a piece only works because it's attached to a blockchain, the art itself isn't doing enough work.
Why the card format is harder than it looks
The trading card aesthetic has a specific set of challenges. The aspect ratio is constrained. There's usually a frame, a title area, stats or descriptors - all of which need to integrate with the character art rather than just sit on top of it. The typography has to work at the final render resolution without becoming a distraction from the central image.
This is where graphic design sensibility becomes as important as 3D skill. A beautifully rendered character inside a poorly designed card frame looks worse than a simpler render in a well-considered layout. The two disciplines have to work together - and that means thinking about the whole composition from the start, not applying the graphic design layer as an afterthought at the end.
What makes a collection coherent
A single strong piece of 3D art is one thing. A collection is a different problem. The cards need to feel like they belong together - same visual language, consistent production quality, a thread that connects them conceptually - while each piece needs enough individual identity to be interesting on its own.
The archetype framework helped here. Because Alien, Human, and Robot are clearly distinct concepts, there is no risk of the collection feeling repetitive. But the production approach - same card format, same lighting logic, same level of finish - creates the visual cohesion that makes it a collection rather than three unrelated images.
The broader lesson
What the project confirmed for me: the craft questions in NFT art are the same craft questions in any other form of 3D character work. How does the character read at distance? Is the lighting doing what it needs to do? Does the composition guide the eye correctly? Is the colour palette considered?
The blockchain layer is infrastructure. It handles provenance and ownership. It does not make bad art good, and it does not make good art less good. The work either stands up on its own terms or it doesn't - and that question has nothing to do with which chain it lives on.
You can see the finished collection in my 3D projects. If you are working on a digital art collection - whether for NFT distribution, branded collectibles, or any other format - the production process is the same. What changes is the destination.