Why don't you bake diffuse? And, normal from bump (baking)?

mortalhuman

1: of all the tutorials and things I learn from, I see people baking AO, baking nomals, but never do I see them slapping on tiled textures intuitively and then baking out a basic diffuse... why not? Seems so much smarter and faster to me than just painting under an AO or pasting textures there... why not bake the whole thing at once and call it Done other than maybe some post editing for polish (perhaps remove seams - don't need to be tiled textures)? :thumbup:

To clarify, you would put your base textures directly onto the mesh, bake it out same time as AO and normal, you're done and ready to go to PS, without all the under-ao pasting and stuff. Why isn't this talked about? I can't be the only person who has come to the conclusion that this is an awesome time saver allowing you to put your library to work, switching things until you're happy, and then burning out the diffuse along with the other maps.

Secondly, I haven't wanted to spend the time to figure it out on my own, so as a bonus question, has anyone ever attempted to bake a normal map from low poly + bumpmaps? Think of the possibilities :x I suppose if it's not possibly, displace, eh?

Opinions on both of these? Is there any reason I don't see anyone raving about the awesomeness of baking diffuse?!

EarthQuake

1. Nothing stopping you from doing a diffuse pass as well. I(and many others) tend to assign RGB values to different sections of the highpoly model, then bake a diffuse map. This lets you easily create masks for the various material types your model may have. Keeping it RGB, or RGBCYMW makes it easy to separate with either a script or by hand(RGB is easy by hand, more complicated is not).

For me personally, I think it wouldn't save me much time creating procedural maps, and definitly would be a waste of time if i had to uv the highpoly, texture it and bake that down.

2. Nothing stopping you from painting a bump map, converting to normals and applying that on your lowpoly. It will never be as good/accurate as a proper normal map baked from a high resolution mesh, and is best suited in combination with a proper bake(ie: bump map for fine details).

cryrid

There are ways of painting diffuse, such as with zbrush or a projection painting program. But the way you're making it sound I'd rather separate them into layers for the same reason I like to composite render passes and work nondestructively; there's no speed lost and it lets me tweak things individually. By being able to paint under the ao, I'm able to tweak the opacity and blend mode of the ao, fix it up and mask it as I deem necessary, while being able to switch out the underlying texture anytime I need to (whether I'm experimenting with difference textures, or if it is just visually getting in the way of me trying to paint in details).

mortalhuman

I love color maps for control

The way I mean is like, it's the same as usual, except instead of laying in separate textures and assembling your base after-the-bake, you'd do the diffuse at the same time, with a bunch of textures (basically same as a color map workflow but with base textures) which bakes out all your textures into the places you'd otherwise need to do all the post-bake psd work for.

You'd have a copied model and the main model, and you'd just basically box map all the elements (or whatever seems best for the part) on the "throw away" model which you will bake down from to your "real uv's" model. Maybe you can do this just by using multiple UV sets on one model, but I've never tried that yet :P

Just wondered because to me, it seems really useful to bake it out, still multiply the ao, and still paint under it, but also to already have that base texture done, with the added benefit you can mix and match bases on elements without having to do switches between the paint app and the modeling app, changing a lot at a time - it's just, drop the new bitmap param and see how it looks - when happy, bake.

My friend says perhaps this will never be as high quality as hand-done, that the renderer isn't capable of the same results. This doesn't make sense to me, but only because of super-real renders, and RTT operations are just renders, so like, is it true? Doing it the psd way is better?

pior

Dont think about it too much and try it!

divi

i tend to map my base textures (basic metal, rust, moss, etc) with blended box mapping to the highpoly and bake it down. the blurriness won't be very visible in the final results, seams on the other hand often would. just saves me an extra step of painting out seams for my base textures.

Orchidface

The method you described only works for an object with unique uvs (non overlapping UVs) which is pretty rare. Usually only characters have unique UVs. The benefit of unique uvs is you can paint everything unique. Baking a tileable texture into a unique UV template is the worst of both worlds- low texel density due to unique uvs and ugly tiled textures. If you want a tileable texture on your prop just use a tileable texture. Anything that uses unique UVs like a shadow map (ambient occlusion) you can use a separate UV channel and a composite material. Almost every prop in current and next gen games has multiple UV channels and composite materials. Sometimes a single prop can have as many as 5 UV channels.

Mark Dygert

I do both, I'll layout a base set of colors to separate areas, TexTools has some good features for this. I also will lay down some base colors using Viewport Canvas on the high and bake that to the low. Its robust enough that I've textured a few low poly models without ever having to crack open photoshop.

Like EQ and Pior said, experiment!