Already baked model, to new topology? Rebaked.

iam717

Is this a bad practice to do?

Basically like lod's but baked down from say 300k to a 100k new version?

BUT using the would've been lowpoly mesh, 300k, to the new 100k final version & NOT baking the highpoly to new version.   Why not? say i couldn't separate the model but the high is separated objects.

Another reason not to do it would be:

Time and it is just faster to rebake already baked maps down to a new mesh, or lod.

Is this how its basically done?

I can see only 1 issue before doing it, which is possible blurring in the textures,(without testing it, when it comes to correct processes or investigations to issues or errors.).  If its useful i came up with it by doing things, if this is done already by some great, I've never heard about it until i did it.

Could some sort of up-scaling & or sharpening sort of fix the blurriness?

I know about not wanting to add things to the normals cause of "warping" or oddities.

I've never done, lod's so i have to assume its new mesh rebake like you did the finer, cleaner, detailed mesh and going down from that but doing the entire rebaking process.

(I've done this before (baked on baked) in 3dsmax, took some time to figure it out, but was fine with the results, others might think otherwise about it, or see things or know things that i do not or have not thought of, which is the point of the thread)

 I can not see wiki cause its on 80/http, 80 is not allowed here, only https.

Eric Chadwick

LODs are usually made for performance purposes. Are you using the 100k as a lower LOD and keeping the 300k as a higher LOD, for switching based in distance or some other metric?

As a side note… less text colors and italicizing would be better. It’s really difficult to read with all the emphases.

iam717

UHM, it was some challenge on artstation i took part of and the limit max was 100k, i agree though i think now a days it is fine but i keep hearing less is best still.  I sort of knew the color situation might be unwelcomed, i agree if it is not me using it i dislike seeing it also. (i was just putting it to use)

The recent character has 9 limbs and can go as high as 11 (spyder-ish design) so the robotic hand details are increasing the count pretty high, plus the stretchy wiggling type arms needing details to retain their curves is also adding to the situation.  For me idc and i unwrapped the 300k version also, but was concerned i'd hear it was too high to use in an engine or game. (wanting the curves to remain.)

Eric Chadwick

Ah ok. If the contest is over, I would recommend keeping the 300k version. Poly counts are much less of an issue these days. And especially not a big deal for portfolio pieces. The biggest thing for portfolios is quality. 

PolyHertz

When you bake from a high poly the generated normal map tries to account for the differences in the two meshs surface depth relative to each other and preserve the way light would interact with the higher poly one. However if you simply transfer a normal map from one mesh to another lower poly one the changes in surface/curvature between the two aren't being taken into account and the way light interacts with it will be wrong (maybe subtle or may be obvious, but it will always be wrong). Best practice is to always bake from the high poly to the actual model the normal map will be displayed on.

All that said, 300K triangles for a character mesh isn't an issue these days (at least on PC/PS5/XBSX , it's a different story if targeting Switch / mobile hardware).

iam717

Thanks guys, yea i did in fact keep the 300k+ , so i essentially made both and was curious about the normal xfers thanks for the information i will NOT be doing that anymore then.