The Dynamic Topology Branch is here! Now in Trunk!


And lot of others.
I was just sitting and waiting for you to say the usual wrong thing.

Well guys, today you learned something from me.
Never, never, use the baking method you suggested!
Use what I suggested, shrinkwrap modifier, even if it isn’t as good as it should. You gonna need to do some more sculpting after but this a very good opportunity to make some more details.
The above baking method (click and shift click etc) produces blurred only results.
Try to bake displacement for instance. Please, use 32 bit (exr)
Then try to displace the subdivided mesh.

Save this and try “my” method using shrinkwrap.
Learn how to use it, see how better maps it bakes from multires.

We had a nice talking, going to bed now. Thanks and Happy blending.

Man, you are so wrong… Shrinkwrap fits one geometry on top of another. It has zero to do with baking. Baking is actual rendering to texture. Shrinkwrapping is using modifier to modify geometry. You won’t get any image output from Shrinkwrap.

The reason the output is blurry is 1. because you use lower res image map 2. Blender Internal renderer sucks.

You want to export your low poly mesh and high poly mesh into OBJ, get xNormal and bake all the maps in the world there. It’s a free app and it comes to Linux next year.

Now it’s your turn to show us “your” method of baking normal maps and displacement maps with Shrinkwrap :wink:

here is a tutorial to bake your normal map in blender

here is some more info

notice the limitations at the bottom.

You know, you’re both correct, depending on your requirements. For stuff like game models where polycounts really matter you want to bake directly like motorsep suggests. For high quality renderings you could technically do the same thing, but it would take more time to retopo to a higher poly basemesh. So you only use the low poly mesh to establish the topology of the final one. Then you subdivide it, shrinkwrap it to the hipoly mesh to make it conform better, apply the subdiv, tweak to fix problems, and bake to that.

For static images, why even bother with baking maps? Just render it as is :slight_smile:

The word “retopo” stands for recreating topology. There is no topology with decimation. No loops that is. It simply reduces number of polygons while trying to keep shape as close as possible. One doesn’t need Shrink wrap for it either. Simply decimate model in MeshLab, bring it back to Blender, UV map it, bake normals using Bias parameter.

When one retopo something, one creates loops. There is no decimation tool that makes loops for artist. And in this case one doesn’t need Shrinkwrap anyway.

xNormal by the way supports ‘cage’, so you can bake any kind of geometry without needing Shrinkwrap.

Foreseeing Michalis saying: " But I want to do it all in Blender!" I can reply: “Then why do you have 3DCoat and ZBrush? And using Blender? Why not just have 1 app?” Specialized tools do job better than all-in-one tools.

  1. Who said anything about the images being static?
  2. Even for static images, surely you want to unwrap, texture and maybe rig your models? Unorganized high poly triangle soups are not very good for that.
  3. I don’t see how cages are remotely relevant. They’re just a method to avoid baking artefacts by altering the normals of the low-poly mesh? Truth be told I’ve never used them, but as I understand it they could be simulated well enough in blender using two low poly meshes - one real and one that bakes nice.

xNormal is really nice though… highly recomended.
http://www.xnormal.net/1.aspx

I just baked a normalmap highpoly DT model to a remeshed base…
(i just did a basic unwrap(marked a couple seams) and did smart uvproject)

baked the normalmap and height map(you can also bake an AO map but that takes some time. i didnt feel like waiting for it… :wink:

well… it works pretty good.

I’m sensing a lot of tension here… :wink:

Not sure I get the problem though.

You can transfer fine detail from an imported messy mesh to your retopo multires by using shrinkwrap and applying it to a high subdiv level of your retopo cage.

The problem is that sometimes shrinkwrap goes wrong and you have to cleanup…

You can also bake selected to active to get details from your messy mesh to your retopo cage and store it as displacement as an exr.
and then use displace modifier (and apply to your top level of multires-ed cage).

The Downside is that blender baking doesn’t compare the limit surface of a subdivided cage so when baking you need to ensure the cage is as highly subdivided as you are likely to go at the time you bake.(which can be slow)

You also need good uvs and a high enough res on the map you bake too… (which Ptex could really help here).

There is no need to apply the multires modifier in either method, but for further sculpting mods
you need to apply the displace or shrinkwrap

The thing is Michalis, i’m pretty sure that you know this so why post like you don’t?

It’d be easier if you posted screenshots ofwhere this process falls short for you. (though there’s already plenty examples of shrinkwrap getting things wrong.)

Posting more concrete examples of what is wrong may help people understand your exact problem and means devs can fix bugs…

If your displacement bake is too soft then try a bigger map or try applying it to a higher subdiv level…
If that is giving artifacts then touch up those areas in the displace map in a paint package, or alter the retopo mesh for a better fit or fix the uvs…

@

It’d be easier if you posted screenshots ofwhere this process falls short for you. (though there’s already plenty examples of shrinkwrap getting things wrong.)

Thanks for replaying Mike. Oh, but I did. Some pages back.
You probably know that most of my posted artwork is based on such technics. Actually, you helped me on this, a lot.

@motorsep
You simply don’t or don’t want to understand what I tried to explain so many times on this forum.

Baking (selected to active method) displacements on a 32 bit map (I wrote this motorsep but you didn’t read it) and displace a multi subdivided mesh, may be a way to have a well organized mesh for more precise further sculpting. You can also have a multires meaning that you pose the mesh if you like. You also have an easier to handle complicated scene (hi poly for renderer only)

Shrink wrap is another method for having a multires mesh. Though it produces some dirt and have to resculpt a bit, it’s my favorite because: it captures crispness from a (here we come in topic) dynatopo mesh better. You don’t have to resculpt everything, just some areas only. After you feel happy you can bake hi to lo poly if you like.

Do you follow motorsep? Sorry for being rude but you insist refusing my thoughts.
My english don’t help me but it’s not the language barrier among us what makes it difficult.

Generally speaking, whatever method we follow, if we bake hi to lo, you might notice some things.
Normal maps are easier to bake, a more forgiving method.
Displacement maps (on 32bit exr) seem blurred, the results as well. Due to BI? I don’t know, probably.
This might be a link to another topic (the “cycles eastern egg…”) where lot of tension exists too. Because normal maps support for cycles is coming soon and there’re geniuses in this forum who believe that bumpmaps are good and normaps are bad.

What I tried to expose in this particular testing thread.
In the past lot of blender users might not face the retopo method as a way for a multires detailed capture.
But lot of other blender users are sculpting on other apps like zbrush. Then, they import their sculpts to blender for further work. These may understand what I’m saying better.
First, because shrinkwrapping is the typical method for baking maps in zbrush.
Second, because zb does it very well and displacement maps are of great quality.

Now we have dynatopology and you may start thinking otherwise my friends.

So much tension! Apologies from my side.
I mentioned shrinkwrapping or “selected to active” methods as a prelude for mentioning something more serious.
Less subdivided dynatopo areas tend to react badly with the above workarounds. You see, projecting a hi subdivided quad mesh on these large triangles captures them. A smooth shader won’t help then. If you look closely on the baked maps you will see these triangles. Shrinkwrap or displacement will still looks triangulated though it isn’t. You have to do sculpting on these areas (smooth at least)

Actually, this might explain why pixologic still refuses to adopt sculptris methods inside zbrush insisting on the dynamesh methods. Think about it.

here is a little test i just did with xNormal and that statue i posted a few pages back…
it started off as a bases mesh from blendswap(Ben Sidmons) i think thats his name… anyhow
u have reuv map the way it was mapped was funky… anyhow…

with XNormal, i was able to bake Normal map and AO map to the base mesh from the hipoly DT mesh.

here is the base mesh wires…
http://img253.imageshack.us/img253/6743/basemale.jpg

here it is Cycles with both maps.
not to shabby for 6000 faces…
http://img189.imageshack.us/img189/9019/normalandaomapxnormal.jpg

The Normal and AO maps were a little funky around the seams…
How i fixed them was i imported the model into Sculptris, when into paint mode,
imported the Texture i wanted to Fix… did my fixing… then re-exported the texture…
it keep your UV layout and everything… Sculptris is nice for texture painting you paint right on the model… press “c” and click the model and it picks that color from that area etc…
no need to export the Obj from sculptris or anything… the UV’s are the same…

Well, I think me and other users misunderstood what you meant, because when you say: “how to capture details from an imported mesh? After retopology on top of it, yes, then what?”

The answer should be: “just bake details from hi poly to Low poly with render to texture
But now i think you’re not interested in producing a map, but rather you want to transfer geometry details onto the new clean mesh, no maps at all. Right?

When you said in every post, “how do I bake details”, you had to specify that you did not want a texture map, but geometry details.

Hope i understood now.

notice that normal map doesn’t seem to be effecting or breaking the silhouette,
working like a bumpmap rather than a displacement map.

@holyenigma74
But… try the same with blender. x normal is an x ternal app. A very good one.
Your tip on sculptris editing. Very clever. I use 3dcoat for such operations.

Normal maps then… but cycles doesn’t support them yet.
I don’t know what you did but they look as ugly bumps, normaps look much better.
You could try BI, but, again, normal maps don’t react with AO. So, it has to be baked too (as you already did)

@marcoG_ita
I spent some time and wrote my thoughts as clear as possible for my english.
You can trust me, LOL, I very well know what we’re facing.

@all
Just for the record, today I compared again two displacement maps on the same figure.
One made by Baking From Multires (mesh coming after shrinkwrap operations) *BFM
A second coming from “Selected To Active” baking operation. *STA

Both BFM and STA are 32 bit exp 2048x2048
After displacement operation on a fresh copy of retopo cage +subsurf modifier:
BFM is a lot crisper and does the job.
STA is for the trashes. Sorry.

So you don’t want to understand why I use shrinkwrap, right? I have a lot of good reasons and a little good taste. Period.

@michalis - post #834

I’d like to understand if i got it.

@ michalis
If you take a look at a dyntopo mesh with smooth shading in the viewport you can see this artefacts too, and if you add subsurface to it, things become even more worse. It’s logical to me, that also rendering / Baking show this kind of artefacts. If you take a look at how meshes look in sculptris, there is maybe hope that a, for dyntopo meshes optimized, mesh smoothing algorithm can fix this problem in the future.

For now I would suggest a different workflow as a workaround:
Use Dyntopo to sulpt out as much detail as you can recreate while doing the retopo job and sculpt all the details with the new quad based mesh. It’s maybe not the most efficient way, especially if you need to do a second retopo mesh at the end to get a nice mesh for animations, but you get finally a good result.

Another, and not tested idea:
Is it possible to bake with cycles? If yes you can maybe try to adopt the shader technic you use for rendering dyntopo meshes to bake your normal or bump maps with cycles.

A third possible workaround could be to remove the baked artifacts from the bump map image in gimp or Photoshop.

yea… that image above is just with a “normal”/bump…
if you want to displace geometry(that will break the silhouette)
bake out a height map with xnormal(blender internal displace at about .02-.025)
you will probably need to a sub-surf modifier on your mesh as well… for the displacement.

@marcoG_ita
You got it. But there’re many reason to transfer geometry on multi res mesh.
This also answers to @Inspir3dArt . The target is a multires mesh with adopted details as crisp as possible. Then, you can sculpt one this, fixing artifacts, adding details. This is why we need dynatopology.
About sculptris. It may look smoother but it isn’t really. Try to import sculptris to zbrush (zapplink) then try there to project a quad retopo mesh. Same problems, even there’s a strong suggestion there, to enable smooth shading on BPR renderer. But it wont help.
You see, it’s the projection (shrinkwrap) method that produces the crisper adoption. Imagine so many quads that snapping on a large triangle… what else can you expect?
@holyenigma74
eh, I noticed that you use normal maps as bumps. Don’t do this, please. I didn’t say about displacement. Neither do this using normal maps. Don’t even try to convert normal maps to bumps (crazybumps app, even gimp). Such tricks may work for you on a stone wall but never on a figure.

sorry i don’t take orders.

@holyenigma74
Sorry, no offenses at all.
Do what makes you happy.
I always reply to your posts, commenting your creations. I didn’t mean to give orders. Who am I?

Less bitching, more sculpting.