Jump to content
3DCoat Forums

AJ's doodley doodles


ajz3d
 Share

Recommended Posts

  • Contributor

Thanks again guys.

 

Wow! That's really cool! If the bullets don't get you, then probably the harpoon will. :)
I like your signature in the corner too!

David Cushman (http://www.dave-cushman.net/shot/ducksfoot.html) wrote that it "is a thin barbed blade that can come into play after the pistol has been discharged". No indication to whether it's fixed in place or not. But I like your idea of a harpoon very much, especially if it could be fired too. :)
 

If you have it it in Voxel mode and reduce the scale, you're counts will go down.

Actually I had to scale my VoxTree up at one point because I ran into some problems that popped up due to the fact that I began sculpting in a far too small scale (camera clipping and brushing issues). Fortunately, when voxel layers contain geometry, they do not increase their density on scale transformations. Otherwise, It would surely clog my CPU. :)
Anyway, I try convert my layers to surfaces as soon as I'm fairly sure I won't be performing any more boolean operations on them.


I did some retopo work yesterday. About half of the model is done, mostly large surfaces like butt, barrels, some brass decorations and half of crocodile. I'd like to keep its polycount around 20,000 triangles and will use three 20482 UV-sets, one for each material: wood, brass and steel. I plan to prepare textures using a combination of 3D Coat's paint room and custom networks in Substance Designer.
 
Here are some initial experiments performed on existing retopo mesh (I hope the image isn't too dark). Texture definition on final brass and steel textures will probably be slightly lower from what can be observed on the image, because there are still many elements sharing these two materials that I have not retopologised yet.
 
retopo_tex_01.jpg?dl=1

  • Like 1
Link to comment
Share on other sites

  • Contributor

Friends,

 

Looks like I ran into a problem...

 

I've always been doing almost exclusively SubDiv models (mostly with retopo shapes already predefined), so doing retopo work for games is something relatively new to me. That said I might not know every caveat a game modeller should know, though I read some stuff at Polycount's Wiki to boost my knowledge a bit in this department. It turns out it wasn't enough.

 

So, the problem.

 

I already retopologised  a lot of my model and most of the objects seem to bake to lo-poly quite okay, however I have trouble with the frizzen. It may seem it has a very simple shape, and there were more complex components of the gun that baked just fine, but with this one I cannot get acceptable results. The normal map of the cylindrical part bakes with ugly radial stripes. There are also some artefacts on the thin side of the mesh when normal map is applied to it.

What might have caused this?

I can upload both hi- and lo-poly meshes for inspection if necessary.

 

ugly_bake.jpg?dl=1

Edited by ajz3d
Link to comment
Share on other sites

  • Contributor

Thank you for the links Carlos, but I don't have problems with UV seams. It's the shading caused by either - the normal map or by my retopo mesh that troubles me.

Edited by ajz3d
Link to comment
Share on other sites

  • Contributor

If I press '6' in the paint room to display the so-called 'med-poly' mesh (which is exactly the same as a lo-poly mesh in this case - because of no subdivision applied to retopo mesh on baking), I get no artifacts at all (at least not those that I should pay attention to). But the normal map 3DC saves to file contains them. So I'm a little bit confused here. What exactly is going on here?

 

hi-med_poly_comparison.jpg?dl=1

Edited by ajz3d
Link to comment
Share on other sites

  • Contributor

the Lo-poly pic (at right) show the uv islands in different colors ?

 

try less edges, in the pivot, in the cylinder

I initially tried with lower amount of edges, but the bake was similar to the last image from the first post of this topic: http://www.polycount.com/forum/showthread.php?t=72713.

Link to comment
Share on other sites

  • Advanced Member

Hard to say.  If I get weird bake errors like that, I either:

 

1)  Bake with different inner/outer cage and bake method

2)  Redo my topology in those areas

3)  Choose a different material in the Voxel Room for the object prior to baking

4)  Use another version of 3DC, I keep the 3.x version always at hand for that very reason

  • Like 1
Link to comment
Share on other sites

  • Contributor

Thanks for suggestions, Robert.

#1 - I tried, but to no avail.

#2 - I'm not sure what else I can do here. The cap of the cylinder that causes me trouble is a flat plane with some spans. What can I do here to further improve the retopo mesh without excessively increasing poly count?

#3 - I use the Default material which always worked in the past. I tried with other materials with the same result.

#4 - I'm not sure if it would work. Especially when xNormal returns me similar crap. I guess it must be something with the retopo mesh.

 

The files are here to download for some kind soul that would like to inspect or experiment with them:

https://www.dropbox.com/s/9r4bi0iopz6dcbl/frizzen_files.7z?dl=1

 

Link to comment
Share on other sites

  • Advanced Member

Are any of those other normals somehow flipped?  I doubt that's it, as they're rendering, just not well.  If not, try adding a span down the middle of one of those problem areas and see if that fixes it in that area.  Otherwise, I'm running out of ideas, other than to take a look at it.

 

One question...how critical is it that you have that particular geometry?

 

EDIT:  Doing my own retopo against that model by hand took about 5 min and baked just fine.  Granted the geometry I created is lower than what you had, as it's often my area to do game-ready pieces, but the bake should work just as well with increased geometry.  I'm going to take a look at it with your retopo mesh next.

 

EDIT:  Yours results in pure black, which means my bake settings are wrong for that retopo mesh.  I adjusted down to 1 (inner) and 3 (outer), but it still wasn't right.  The time it will take to figure out the right settings probably isn't worth it vs retopoing it again.

 

My suggestion is to retopo it again.  I've done it twice and it baked well the first time, the second time I had to adjust a couple of things, but the retopo was aimed at being more like what you had before.

Edited by alvordr
  • Like 1
Link to comment
Share on other sites

Look first image

 

set:

Merge into scene = microverts

 

 

O.K. after a few hours of trial and error I got some fairly good results using a low poly retopo mesh (5k triangles). Using the Baking settings of "Scan inside/outside the object" set to 1.2, using the "Snap to outer surface" method and adding a few zones with their depth set to 2.0 over areas which were not captured, (BTW a mirror support would help a lot with adding zones to a symmetrical mesh). I then used "Merge into scene (microverts)" and selected the lowest Carcass resolution with 20 million for the hi rez map. I had to do quite a bit of smoothing though and the exported displacement map didn't hold all the details as what was in the view port.

 

It seems that the culprit is my UV map, The area where I am getting problems on the mesh is rather reduced on the UV map, so I think I will have to take better care of my UV real-estate, it appears that if you don't have a large enough UV area, when the displacement map is created, you will get a transparent area with floating checkered pattern there instead of a map and no matter what you do you cannot fix this unless you give that area more UV space.

 

and

 

http://3d-coat.com/forum/index.php?showtopic=10707&p=83078

 

:D

  • Like 1
Link to comment
Share on other sites

  • Contributor

Thank you for your time, friends. I appreciate your help.

 

I think the problem is solved, although I did not suspect this kind of a solution.

 

Here's what I did.

 

First, I tried baking normal map from Carlos' scene using the microverts path. But the exported normal map, while looking pretty decent in the paint room, when imported into Substance Designer turned out to be full of artefacts.
post-12523-0-69004300-1407108342_thumb.j

Then, following Robert's advice I tried to add some edge loops and noticed that inserting them very close to inner and outer UV seams of a cylinder completely eliminates those ugly radial shading artefacts. But the cost of this operation was very high because it would add a lot of unnecessary geometry. So I thought that it would be a good idea to follow Robert's other advice - making a new retopo mesh.

I started with a simple plane and noticed that it bakes perfectly. Then I made a flat cylinder's cap and this baked very well too. But once I started to connect both cylinder caps from the outside, artefacts began to emerge on bakes. When I connected the inside, the mess was complete - the very same artefacts appeared that I had on bakes from my original retopo mesh.

Then I recalled some article I once read that explained how to fix shading errors on a game mesh. It was something to do with manually triangulating offending faces. So I triangulated all faces of cylinder's cap and voila, the cap baked without any of those friggin stripes.

post-12523-0-88691600-1407110088_thumb.j

And the best thing is, that with the newest beta version of 3D Coat (4.1.11A) I don't even have to manually triangulate the mesh before baking because Andrew added support for triangulation. This, from what I understand, triangulates the retopo mesh only for the time of baking and according to algorithm selected by user. "Simplest" triangulation works very well with Substance Designer (it's intended to be used with Unity). I didn't try Delaunay with SD yet, but in 3D Coat it bakes perfectly too.

post-12523-0-26344300-1407110140_thumb.j
 

Edited by ajz3d
  • Like 2
Link to comment
Share on other sites

  • Contributor

Actually, I just took another look at the NM PPP baking menu and there's an option that enforces triangulation, which I did not enable when I was baking. So most likely I was completely wrong praising this new function, because it didn't have anything to do with the successful bake as it was simply turned off. Also, it permanently triangulates the mesh.

So, if it did not, then what made my bake successful with the original retopo mesh? Maybe it's the Tangent space standard? I don't know.

 

Ignore the above. Turns out I had right in my previous post. I did some tests and triangulation method DOES seem to have significant impact on the bake result even when Triangulate from the baking menu is disabled. Delaunay's algorithm causes artefacts on my object, Simplest does not.

I'll post some images for comparison later.

Edited by ajz3d
  • Like 1
Link to comment
Share on other sites

  • Reputable Contributor

I downloaded your posted files, the first ones you gave in the thread.

3DC version Windows 4.1-11A non cuda.

Merged the high polygon frizzen mesh into surface mode.

Imported the low polygon frizzen mesh as the retopo mesh.

Left the Uv set as is.

Chose 2k image resolution.

Merged to paint room using PPP with no subdivision plus no triangulation.

 

The model seems to have the correct normal map in the paint room with no artifacts.

post-518-0-99395300-1407191426_thumb.png

Edited by digman
Link to comment
Share on other sites

  • Contributor

Yup, the newest version is capable of baking this particular model really good. Especially if Simplest triangulation is selected as triangulation method. But it does have its share of problems when baking some other of my pistol's hard surface elements, even if retopo meshes are carefully modelled, with UV seams and hard edges in appropriate places. It presents slightly inferior quality to what xNormal is capable of (when using custom edited cages instead of inner/outer ray distance):

3dc_xnormal_bake_comparison.jpg?dl=1

 

As a side note, baking with xNormal comes with a trade off of course, because there's no such thing as Use Names Correspondence for Baking or something else that would automatically explode the mesh and bake the map. So this slows down the pipeline a lot, unless the program in which we explode our object is capable of scripting and can take on huge reference meshes without significant impact on viewport performance and time needed to load/save references. When it comes to the scripting part, in Blender or Maya for example, it should be possible to script in export routines for reference, retopo and cage meshes that would automate the process and reduce artist's time spent on doing those actions.

I'd also like to mention two problems that I have not yet reported to Mantis.

The first one is that if I bake my meshes with locked normals, I get this thing in the viewport after I press '1' to view the relief map:

post-12523-0-28349900-1407355652_thumb.j

This mess doesn't affect the exported normal map, but it makes painting additional normal map details a little bit... um... difficult.

The second problem is more bothering. After baking I cannot paint anything on my mesh unless I disable Ignore back faces. This behaviour suggests that the normals of my retopo groups are inverted, but they aren't! And this I checked not only in 3DC, but in 3rd party programs as well! Also, 3DC's backface culling won't hide them so WTF? If I rotate baked model in the paint room, I get black faces because of this.

My frustration level with 3D Coat has risen by a considerable amount. :girl_devil:

Why on Earth it's so difficult to go through the pipeline with this program?

Edited by ajz3d
Link to comment
Share on other sites

http://3d-coat.com/forum/index.php?showtopic=16476&p=111699

 

... every external package may treat the quadrangulation differently when importing the model which will result different normal information on the model and then the normal maps wont display correctly.

 

In short  -

 

for the normal maps to display correctly the model should be triangulated so the normals wont change .The external package must import model with the normal information  and the correct tangent basis should be used for the external package to display it correctly.

Link to comment
Share on other sites

  • Contributor
This doesn't explain seams in bakes done by 3DC and visible on the first image of my last post. For the comparison image I posted (screenshot from Maya) I chose the Delaunay triangulation, which is said to be close to what Maya uses. Also, I used "Maya normals" as normals calculation method and "Maya, Blender" as normal maps export. Tangent space standard didn't matter much as it only slightly affected shading, not seams.

Triangulating the model makes it hard to introduce modifications into it later on. That's why I only triangulate faces that might be ambiguously interpreted by an engine's triangulation routines. Convex? Concave? I don't want to leave this decision to algorithms.

I don't expect magic resulting in no seams and artefacts at all, but if some software can bake better normal maps using the very same retopo mesh than the other one, then maybe there's still some room for improvement in the latter?

Edited by ajz3d
Link to comment
Share on other sites

  • Reputable Contributor

I think what we need to improve the merging (baking) into the paint room is the ability to set each retopo group cage per voxel / surface layer. We can use sphere of influence of course but it would be must easier to set each retopo group cage per voxel / surface mode layer when using Name Correspondence. Since Name Correspondence only bakes one layer at time this would be ideal. Then you could use sphere of influence if necessary on each retopo cage layer if needed.

 

Right now you have to just choose a general fit over the entire model which is not the best for baking when using Name Correspondence. It works fine on a one layer object but when you have many layers some areas are too close to the cage an other areas are too far away. You have to just set a best general fit so all areas are covered and using sphere of influence on a many layer complicated model is very time consuming.

Edited by digman
  • Like 2
Link to comment
Share on other sites

  • Contributor

My frustration level with 3D Coat has risen by a considerable amount. :girl_devil:

Why on Earth it's so difficult to go through the pipeline with this program?

 

Don't bother. Everytime I tried to bake in 3dc I got weird shading, wiggly lines even by playing with the area of influence and other baking setting, while on the other hand, exporting everything and baking in xnormal is a breeze.

I'm not a fan of hammering a project which should move pretty quickly because the software is doing unexpected thing. There may be a certain WAY of doing things in 3dcoat which result in good output. Unfortunately I didn't find it and to be honest: if I've to guess how to do something and know the theory, I won't use a software that keep getting in my way because IT wants to do it his way ;)

Results talk, processes don't matter.

Link to comment
Share on other sites

  • Contributor

digman, on 07 Aug 2014 - 01:42 AM, said:

I think what we need to improve the merging (baking) into the paint room is the ability to set each retopo group cage per voxel / surface layer. We can use sphere of influence of course but it would be must easier to set each retopo group cage per voxel / surface mode layer when using Name Correspondence. Since Name Correspondence only bakes one layer at time this would be ideal. Then you could use sphere of influence if necessary on each retopo cage layer if needed.

 

Right now you have to just choose a general fit over the entire model which is not the best for baking when using Name Correspondence. It works fine on a one layer object but when you have many layers some areas are too close to the cage an other areas are too far away. You have to just set a best general fit so all areas are covered and using sphere of influence on a many layer complicated model is very time consuming.

I wholeheartedly agree with all you said here Digman. General settings of bake scan depths don't play well with name correspondence baking. Every object should use its own scan depths. Sphere of influences should also be bound to objects instead of influencing every scan depths of every single object within their radius. This would make things much easier than trying to pick this perfect depth that would cover all objects and then, having to compensate with dozens of influence spheres where needed.

 

alvordr, on 07 Aug 2014 - 04:56 AM, said:

I don't know.  I've had almost no real troubles with baking, and I never use the spheres.

I'm glad to hear this, Robert. Unfortunately I cannot say the same. Generally, most of the bakes I've been doing in 3D Coat did in fact bake well, but they were mostly SubDiv stuff. However baking retopo meshes that are not meant to be subdivided, like the duckfoot's pistol here, is a lottery for me in 3D Coat. Some bake well, some contain artefacts.

xNormal on the other hand bakes those meshes perfectly. :pardon:

Maybe you could share a tip or two on how to bake game stuff in 3D coat? It seems to be full of caveats.

 

BeatKitano, on 07 Aug 2014 - 2:51 PM, said:

Don't bother. Everytime I tried to bake in 3dc I got weird shading, wiggly lines even by playing with the area of influence and other baking setting, while on the other hand, exporting everything and baking in xnormal is a breeze.

I'm not a fan of hammering a project which should move pretty quickly because the software is doing unexpected thing. There may be a certain WAY of doing things in 3dcoat which result in good output. Unfortunately I didn't find it and to be honest: if I've to guess how to do something and know the theory, I won't use a software that keep getting in my way because IT wants to do it his way ;)

Results talk, processes don't matter.

Ya, I'll stick with xNormal for now. Probably until I'll feel that I'm ready to do some more experimenting with 3D Coat's baking engine.

Frankly speaking, organic models seem to bake very nicely in 3DC, but whenever I try to bake a hard surface mesh, the results are unpredictable. Sometimes very good, sometimes not. The more rough the mesh is, the worse the outcome. Even if I lay down hard edges and lock the normals.

Maybe cages are a superior solution to bake scan ranges?

 

Also, 3DC as far as I know 3D Coat still doesn't support normal map anti-aliasing, and xNormal does.

---

Today's quick doodle. ~200k tris so far:

doodle_oldMansHead.jpg?dl=1

  • Like 2
Link to comment
Share on other sites

  • 2 months later...
  • Contributor

I had some free time yesterday, that I spent working on textures of my ducksfoot pistol.

A mixture of 3D Coat and Substance Designer. Both programs complement each other very nicely.

duckfoot_pistol_butt_wip.jpg?dl=1

Edited by ajz3d
  • Like 4
Link to comment
Share on other sites

  • Advanced Member

Nicely done!  I have SD, but I don't use it, so I don't really know how to use it.  I tried a couple of times and felt that my current programs allowed me to do what I needed to, just fine.  However, if you feel SD is worth learning, I might have a go at it again.

Edited by alvordr
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

×
×
  • Create New...