Baking on a cylinder not working correctly?

TannedBatman

Whats going on here? 


Shouldnt the averaged normals stop this? Or am I forgetting something.

gnoop

Based on your vertex normal screen  they are all  split  "explicit" normals in therms of 3d MAx   doing gazillion "hard" edges.  Perhaps  it's something with that.  But in general  baking some complex shape into a simple cylindrical form  would never be  looking right  without  proper control of projection direction  by  a cage  and probably some edge flow projected on the cylinder  too.

I would rather do like 6 edge  cylinder but with that Easter island face  in  the geometry. At least nose and  face plane .

TannedBatman

gnoop said:

Based on your vertex normal screen  they are all  split  "explicit" normals in therms of 3d MAx   doing gazillion "hard" edges.  Perhaps  it's something with that.  But in general  baking some complex shape into a simple cylindrical form  would never be  looking right  without  proper control of projection direction  by  a cage  and probably some edge flow projected on the cylinder  too.

Odd, im certain these arent hard edges as I unlocked the normals and pressed soften edge.  Which is why im confused, il have to experiment with a cage.

gnoop

TannedBatman said:

Odd, im certain these arent hard edges as I unlocked the normals and pressed soften edge.  Which is why im confused, il have to experiment with a cage.

It's 3d max, right?   I clearly see in your screen that  it's two normals  shooting  out of same vertex point . 

TannedBatman

gnoop said:

TannedBatman said:

Odd, im certain these arent hard edges as I unlocked the normals and pressed soften edge.  Which is why im confused, il have to experiment with a cage.

It's 3d max, right?   I clearly see in your screen that  it's two normals  shooting  out of same vertex point . 

Its maya, and yeah the screenshot shows hardened edges I just realised, but ive tried softened and its the same. I think the projection is off because the nose is too deep and it projects it on a surface much further out therefore it stretches. I still dont know how to get what I want outside of snipping together normal maps. I cant quite get a cage to fix it either.

pior

Your images do a very poor job at explaining the issue you are running into.

"Shouldnt the averaged normals stop this?"

What is the "this" you are hoping to stop ?







............. Also ... why in the actual F do you have a photo of Jake Gyllenhaal as your avatar ?

sacboi

pior said:

............. Also ... why in the actual F do you have a photo of Jake Gyllenhaal as your avatar ?

(was wondering about that too?!)

Not often your posts prompt an attack of the giggles

as to the issue at hand, OP if you can further clarification please!

TannedBatman

sacboi said:

pior said:

............. Also ... why in the actual F do you have a photo of Jake Gyllenhaal as your avatar ?

(was wondering about that too?!)

Not often your posts prompt an attack of the giggles

as to the issue at hand, OP if you can further clarification please!

lol, idk I just like the movie, should I have something else? Anyway the issue is the high poly is being projected incorrectly. You can see the dorsum of the nose is being spread wide. I gathered it is because of how deep the nose is there in comparison to the low poly so it stretches out, but it also obviously doesnt know which angle it is to be viewed from so it generalises it I think. Not sure how to control that though whilst also getting a clean bake elsewhere, maybe if I manually adjust the normals near the face?

pior

A few things you could do : 

1 - provide your low and high, for people to be able to actually help you track what's going on.

2 -
"You can see the dorsum of the nose is being spread wide."
"the high poly is being projected incorrectly."
These sound like very subjective statements. Are these things actually incorrect ? One proper investigative approach would consist of visualizing a cross section of the models and draw the expected rays as one would expect them to be cast. That way you'll be able to tell for certain that something is not as you would expect.

But since you are not providing the models, people can't investigate this along with you. Also I stand by my remark that the screenshots provided are poor. You are tilting the models for no reason, and so on. Take the time to properly present your issue, putting in as much care possible.

gnoop

Your low poly cylinder have so dense geometry  you could use to make  those nose and face  instead .    There is a general misconception  that  cylinders need more  facets to have nice  shading and bakes.      Could be just 6 facets  in your cylinder but still with that nose modeled in .     

If it's a mobile game and you need as low poly as possible   do just  triangle profile for that cylinder   with the nose coinciding with one of the edges . it would have more chances  probably.

TannedBatman

pior said:

A few things you could do : 

1 - provide your low and high, for people to be able to actually help you track what's going on.

2 -
"You can see the dorsum of the nose is being spread wide."
"the high poly is being projected incorrectly."
These sound like very subjective statements. Are these things actually incorrect ? One proper investigative approach would consist of visualizing a cross section of the models and draw the expected rays as one would expect them to be cast. That way you'll be able to tell for certain that something is not as you would expect.

But since you are not providing the models, people can't investigate this along with you. Also I stand by my remark that the screenshots provided are poor. You are tilting the models for no reason, and so on. Take the time to properly present your issue, putting in as much care possible.

Sorry, the reason I didnt provide the files was because Its basically just a cylinder, and I thought the issue was fairly obvious, the photos appear tilted because the model is horizontal and tilted. Regardless, ive cleaned up my low poly, positioned it upright and taken new screenshots and provided files. I dont think the dorsum being spread wide is a subjective statement, if you look at the high poly and then the bake its very clearly the result. The rays are doing exactly as I would expected, but not what I want. I said in a previous reply I believe the depth is the issue and its spreading the details over a large gap.



What I achieved by using a perspective bake, by using a miniaturised version of the low poly as a cage facing the direction of the model. However you can see this messes up the rest of the model. I hope I attatched the files correctly, I just now realised theyre fbx and should probably be obj, not sure it matter t

TannedBatman

gnoop said:

Your low poly cylinder have so dense geometry  you could use to make  those nose and face  instead .    There is a general misconception  that  cylinders need more  facets to have nice  shading and bakes.      Could be just 6 facets  in your cylinder but still with that nose modeled in .     

If it's a mobile game and you need as low poly as possible   do just  triangle profile for that cylinder   with the nose coinciding with one of the edges . it would have more chances  probably.

You're right, I could easily achieve this with higher geometry in the low poly, but I wanted to test first with a cylinder, even if its not what I go with in the end I would still like to learn how to solve this issue in case I come across it in the future.

pior

Red is your lowpoly cylinder, highlighted for emphasis.
Blue is the directions of the rays, cast along the normals of the surface of the lowpoly
White is where detail gets sourced and projected.

And as suggested above, the models are shown at a cross section showing the details you are not happy with.


Everything is behaving exactly as it should - to use the wording in the thread title, it is "working correctly".

Now that the above is established, what is it that you want to achieve ? What are your *actual* constraints ?

The thread makes it sound like you absolutely have to use a freeform head sculpt as a source, and that you absolutely have to bake to a cylinder ... even though in practice your are obliged to neither. So which one is it, Donnie ?

TannedBatman

pior said:

Red is your lowpoly cylinder, highlighted for emphasis.
Blue is the directions of the rays, cast along the normals of the surface of the lowpoly
White is where detail gets sourced and projected.

And as suggested above, the models are shown at a cross section showing the details you are not happy with.


Everything is behaving exactly as it should - to use the wording in the thread title, it is "working correctly".

Now that the above is established, what is it that you want to achieve ? What are your *actual* constraints ?

The thread makes it sound like you absolutely have to use a freeform head sculpt as a source, and that you absolutely have to bake to a cylinder ... even though in practice your are obliged to neither. So which one is it, Donnie ?

I feel like we established the aim a while ago, I have also explained twice the reason things are behaving correctly and as expected. The title was meant less in a way of "correct" or more in a way of "desired", since making the title and first comment my knowledge of the situation has grown. Originally I thought what was expected was different but upon thinking some more the result was exactly what would have been predicted. My comment before this shows with my perspective bake the aim I want to achieve. I want a flat front view of the face with no stretching on the nose, but with the rest of the model more or less how it is. I thought what I want to achieve was fairly obvious once you look at the high poly and the bake result, im not saying the result isnt expected however, maybe at first I did though. Im not necessarily obligated to use a freeform head sculpt as the source but im not sure I understand what you suggest as an alternative. I would like to bake to a cylinder for the sake of experimentation and learning if and how it is possible, because I thought that it would. I think the only solution is changing the normals near the face to point toward it, and the rest left as they are.

pior

Well, there are two solutions.

The first one consists of not working on the source highpoly in a vacuum, but rather, considering it for exactly what it is : a source that surface details are being taken from. Therefore, *if* the constraint is that you absolutely have to bake to a cylinder, then the straightforward way is to treat your high as a bas-relief sculpted on a cylinder similarly to a Tiki mask - as opposed to wasting time on a freeform sculpt.

The second solution consists of matching the lowpoly to the volume of the high. That's all there is to it really.

But ... if this is just "for the sake of experimentation" and there are no actual constraints (you actually don't *have* to use a freeform sculpt, and you don't *have* to bake to a cylinder either, and these are just arbitrary constraints chosen for the sake of it), then I would say that the experiment is concluded as you've seen the limit at which things break. And you can now apply this knowledge of what to avoid to your next projects.

gnoop

TannedBatman said:

 I want a flat front view of the face with no stretching on the nose, but with the rest of the model more or less how it is.

To achieve this you can  edit  the cage .  I am not sure about  Maya but in MAx and Blender   the tracing is happening  in direction  between low poly vertex and corresponding  cage vertex.      You can  make the  "face" side of a cylinder doing more like planar  projecting  , lesser of a lens effect  for an expense of  stronger distortion at  sides of the face.  It's hard to say how ugly it may be there  but you could always  find some trade off.    Sometimes it's just very subtle cage touches.  

 Expand you cage more , it's easier  to edit   that way.    

ps. It may takes  hours of miss and hit. It's why it's so hard to do a decent bake for a mobile game usually.     You can also edit the normals itself  but I suggest to start with the cage first.       And the last resort is  sliding  vertexes post-bake.  Usually its a blasphemy  ha-ha  but   sometimes helps for  super low poly things.

   

TannedBatman

pior said:

Well, there are two solutions.

The first one consists of not working on the source highpoly in a vacuum, but rather, considering it for exactly what it is : a source that surface details are being taken from. Therefore, *if* the constraint is that you absolutely have to bake to a cylinder, then the straightforward way is to treat your high as a bas-relief sculpted on a cylinder similarly to a Tiki mask - as opposed to wasting time on a freeform sculpt.

The second solution consists of matching the lowpoly to the volume of the high. That's all there is to it really.

But ... if this is just "for the sake of experimentation" and there are no actual constraints (you actually don't *have* to use a freeform sculpt, and you don't *have* to bake to a cylinder either, and these are just arbitrary constraints chosen for the sake of it), then I would say that the experiment is concluded as you've seen the limit at which things break. And you can now apply this knowledge of what to avoid to your next projects.

I agree, you raise a valid point about doing it like a tiki totem which is basically exactly what this is, I didnt think of that. I might aswell use a cylinder as a base and sculpt on that and il get exactly what I asked for. Problem solved, gee that got way too complicated lol. Thank you for the help 👍
I just realised quote doesnt mean reply and I can just leave a comment and it comes up as a notification.