Thread by @delaneykingrox

Got a DM question about how to maintain texel density when unwrapping game meshes.

I unwrap using Maya LT at the moment, and it has a feature called get and set under the scale transform of the uv tools.

Got "get" the texel density of a polygon, and the "set" the others.

Maintaining texel density is most important at seams, where it becomes very clear when there is a mismatch between the two.

However the actual truth is you don't have to maintain a perfect texel density across all of a model. Some parts just don't get seen by the player...

For example who gives a rusty old damn if the inside of a character's mouth, guns inside barrel or the underside of cars mudflaps are lower res than everything else?

You can often cheat density to allow key features close to the camera are crip.

So part of the question was when to know to split off to a second texture. Well, this comes from the complexity of the model, and any modularity that may be needed.

Say I was doing a space marine from 40k. It makes sense that the majority of the body armour...

Is on one texture. However the pauldrons, insignia and head vary from character to character.

The smart move would be to have the head and neck area on one texture, so it can be super crisp and detailed for cinematics and the skin would have sub surface maps the armour doesn't.

Likewise, if I had an emissive (glowing stuff) cybernetic eye and a light here and there, but the whole armour has no emission, it saves a big empty texture to separate those parts off.

The general rule to break a texture up into several is that you cannot...

...get the resolution you need on all the parts within the max texture size (usually 1024x1024 or 2048x2048 depending on the project platform).

You can figure this out using screen appearance size. Say we do a huge closeup on our hero's face and we want to do this in HD 1080p

We want at least 1080 (the vertical screen size) for the face texture, and higher if possible because to describe pores needs a few pixels.

(You can achieve this by layering masked tiled fine detail textures in the shader -however, but that is for another lesson)

So it makes sense to have a 2k face texture.

But if your hero gets only close enough for a quarter of the screen, then you are looking good at 512x512.

So back to laying out UVs, I start by laying out the biggest, most important islands first and using "get" to see..

...what kind of texel density they have... then I picture that in my head and think about the output size. If I am happy, I will cram in everything else to that density.

Sometimes, however, I waste texture space NOT cramming in the UV tightly, but ensuring the shells are...

Aligned and scaled in a certain way. This allows me to apply effects such as offsetting uvs for damage variants, overlaying wood grain that runs a certain direction and so forth.

This is why I push students to learn to unwrap manually, and not rely on packing algorithms all the time. Knowing how to uv map well allows me to do a tonne of tricks.

And that brings me to the third part of the DM question... substance not working well with tiled overlaps...

So, say I have a character with loads of spikes on their armour. I could bake say, three or four of the spikes and then reuse the texture on all the rest by overlapping UVs.

But your bake will fuck up if you do this.

The answer is you just export two versions of your model...

One has the four spikes unwrapped, whilst the other spikes have their UVs offset off the 0-1 uv square,or scaled down and placed on an empty bit.

When you bake, these polys won't overwrite the four you want.

You then switch the model for one with overlapped uvs to paint it.

But, I wanna just add that the old school way we did texel density us still with us.

Checkers.

That black and white checks texture lets you eyeball the density.

Works a treat unless you need absolute precision :)