Fur is a value problem, not a texture problem
For about four years I drew fur by drawing hairs. Thousands of them, tapered, layered, following the flow of the form. The results were technically impressive at full zoom and looked like carpet from across the room.
What fixed it was not a better brush. It was accepting that fur is mostly an edge and value problem, and that texture is a garnish applied to roughly four percent of the surface.
Three zones, not one surface
Any furred form divides into three regions that behave completely differently under light. Treating them as one continuous texture is the mistake that produces carpet.
The compressed core is not visually fur at all. Hair lying flat against a form under any reasonable lighting behaves like a slightly matte solid. Render individual strands there and you are adding high-frequency noise to exactly the region that is supposed to carry the large shapes.
The transition zone is where shading happens, and it is a value problem with a soft edge. Nothing else. No hair detail belongs here at all, which is counterintuitive, because this is the region that occupies most of the canvas and therefore feels like it deserves the most work.
The fringe — the silhouette edge where hair leaves the form and catches light against the background — is the only place that reads as fur to a viewer, and it is a thin band. All the texture lives there. Detail spent anywhere else reads as noise.
The brush size test
Whatever brush size you think you need for fur, halve it, then use it a quarter as often. Almost every fur rendering problem I have diagnosed in my own work came from strokes that were simultaneously too large and too numerous.
If the texture is visible in a thumbnail, it is too strong. Fur should appear when the viewer steps closer, not before.
There is a related failure where the strokes are the right size but all the same size, which produces a uniform field that reads as fabric. Real fur clumps. Three or four strands stick together, then a gap, then six more. That irregularity is doing more work than the strand count ever did.
Layer count
I used to work on eleven or twelve layers per character. Base, shadow, occlusion, two rim passes, fur underlayer, fur overlayer, and so on. The piece would look progressively worse with each addition and I would add another one to compensate.
Current setup is three: flats, one shading layer in multiply, one light layer in screen. Texture goes directly onto the flats, destructively.
The discipline of not being able to undo the texture pass forces the decision about where texture belongs to happen before making it, which turns out to be the entire problem. Eleven layers let me postpone that decision indefinitely, and postponing it is why the pieces looked like carpet.
What this does not solve
Long fur, feathering and manes are a different animal. There the fringe zone is large enough to become the dominant shape, and silhouette rules take over from surface rules. Short-coat species are where the three-zone model is most useful, and where I now apply it without thinking about it.
It also says nothing about colour. Fur that is correctly structured in value and wrong in temperature still looks like plastic, and noticing that took another year. There is a version of this note about temperature that I have started four times and abandoned four times, because I do not yet understand the thing well enough to compress it.