Two ways exist to make fabric behave. You can run a solver that works out where every vertex should go given gravity, collision and stiffness, or you can shape the cloth yourself and let it stay where you put it. Both are used constantly in production, often on the same character, and the choice is about control rather than about which is more correct.

What simulation is good at

Solvers are extremely good at things that are tedious and unarguable. Fabric draping over a chair. A sheet responding to a body sitting on it. A hem that swings correctly when someone turns. These are cases where a human would spend hours and still get it slightly wrong, because folds obey rules our eyes recognise but our hands do not reproduce.

They are also good at contact. When two bodies press against fabric between them, the solver handles the interaction without anyone deciding on it. That is worth a great deal in close scenes, where the alternative is an artist manually pushing vertices around a contact zone and producing shapes that look almost right and therefore look wrong.

What simulation is bad at

It is bad at doing what you want. A solver produces a plausible result, not the result that composes well, and there is no slider for make this fold read better in silhouette. The usual workaround is to cheat the inputs: add invisible collision objects, adjust wind, or animate stiffness over time until the physics happens to produce the shape you wanted.

It is also slow and fragile. Simulations explode when geometry interpenetrates, when a character moves faster than the substeps can follow, or when a collision mesh has a hole in it. Every production has lost a day to a skirt that turned into a knot on frame ninety-two, and the fix is usually to reduce ambition rather than to increase settings.

Hair is a different problem

Real hair is hundreds of thousands of strands with complex self-collision, and simulating it accurately is expensive even for large studios. Stylised work almost never wants the real answer anyway, because a drawn style reads hair as a few large masses with clean edges, not as a cloud of individual fibres.

So most stylised hair is modelled as geometry: clumps, ribbons or cards, rigged with a simple joint chain and animated or lightly simulated as a rigid structure. That gives clean silhouettes, predictable shading, and the glossy band that reads as anime hair. Simulating the same thing accurately would produce a noisier and less attractive result at ten times the cost.

Cache early, cache often

A simulation that is not cached is a simulation that will change. Solvers are sensitive to frame ranges, substep counts and start states, and a scene reopened tomorrow can produce a different result from the same file. Cache to disk, name the cache, and treat it as the real asset. The simulation setup is a recipe, the cache is the food.

Cache before lighting, not after. Lighting decisions get made against specific fold shapes, and rerunning the sim afterwards moves those shapes and invalidates the work. This ordering problem is the same one described in the blocking piece: expensive stages should depend on locked cheap ones, never the reverse.

Substep count is the setting most worth understanding. It tells the solver how many times to check collisions between one frame and the next. Low counts are fast and let fast-moving geometry pass straight through a body; high counts are slow and stable. Raising substeps fixes more problems than raising any other value, and it is the first thing to try when a sim misbehaves rather than the last.

Combining simulation and hand work

The practical pattern is simulate, cache, then sculpt on top. Run the solver to get the base drape, freeze it, and then hand-adjust the handful of folds that matter for composition. That gives you physics where physics is tedious and authorship where authorship matters, which is the right division of labour.

Keep the characters fictional and unmistakably adult, and remember that clothing is part of how age reads in a frame, not a separate decoration. The palette side of fabric is covered in the colour notes, and the proportion side in the stylised proportion piece.