Why bake down normals from a highpoly when you can convert your diffuse into normals?

poopstick

Ok I'm really confused, I recently found that you can convert a diffuse map into greyscale and then convert that into a normal map. This seems too easy, but it does seem to work most of the time.

Is this what people do in the industry or should you always bake from a higher poly?

Obscura

Well, it looks like your confusion is not about the workflow but the meaning of normalmap itself. Read this:
http://wiki.polycount.com/wiki/Normal_map

Farfarer

Generally covering a diffuse map into a normal map isn't going to give you good results.

By and large the best that can be done is it treats the diffuse map as a height map... so anything that's bright is raised up and anything that's dark is sunken down.

If you look at many diffuse textures in that way, you should be able to easily tell that's not often the shapes that the diffuse map is representing. Especially if there's any lighting in your diffuse map - the bits in highlight will be higher than the bits that are in shade... even if it's a flat surface.

But if you bake all that from a high poly mesh, you know what you're baking it's actually a true representation of the surface you want.


As for what's done in the industry... it's variable. Sometimes it's just easier to convert a diffuse map to a normal map, but generally you'll want to give it a bit of work first to ensure its actually a half-decent height map. It'll do in a pinch, basically, or for stuff that's not too important.

a3D

Baking is irreplaceable if you're doing normalmap-based bevels and fillets (in order to make a border edge look smooth when the existing geometry wouldn't allow it).
You will never guess the tangent right if you try to imitate that with a handpainted map.

On the other hand, adding surface detail (like wood grain, fabric small reliefs i.e. screws) can and should be done by converting pictures or patterns into normal maps, using tools like Crazybump or Quixel's NDO2.