Bending All The Wrong Places: XNALara to SFM Port via Blender (Troubleshooting)

[SUPREME]

40 second video below illustrating the issue (also included Blender footage so you can see armature). I'm fairly new to porting so please assume I don't know much. The rig (.py file) I'm using with the model was a basic setup created using the Auto Rigger tool following a simple connection of pelvis > upper leg > lower leg > foot. Mostly working but the model is bending in weird places in the leg.

Any thoughts? Been working on this model for two days :(

https://youtu.be/ALzqOQtotjE

[Pte Jack]

The problem you have here is most likely a weight culling problem. When you first exported your DMX files you most likely found that you had a lot of verts that had weight influences greater than 3 bones. You culled them out. The weights culled out had to be distributed back into the chain of bones so this caused a problem.
OR
You did not set up the Autorigger correctly and the wrong bones are being used...

It may be those 2 butt bones that are giving you the problems if you didn't make them jiggles. You've also got a lot of bones up in the shoulder area. These may be twist bones but it looks like they maybe parented wrong into the upper arms and are being missed in the rig. Hard to say without actually touching the model.

I would check the autorigger set up first.

f that isn't the problem, I would decompie the new model, bring that back to Blender and chech the weight painting and make adjustments where necessary to blend them over the 3 bone limit better.

[SUPREME]

Originally posted by Pte Jack:

The problem you have here is most likely a weight culling problem. When you first exported your DMX files you most likely found that you had a lot of verts that had weight influences greater than 3 bones. You culled them out. The weights culled out had to be distributed back into the chain of bones so this caused a problem.
OR
You did not set up the Autorigger correctly and the wrong bones are being used...

It may be those 2 butt bones that are giving you the problems if you didn't make them jiggles. You've also got a lot of bones up in the shoulder area. These may be twist bones but it looks like they maybe parented wrong into the upper arms and are being missed in the rig. Hard to say without actually touching the model.

I would check the autorigger set up first.

f that isn't the problem, I would decompie the new model, bring that back to Blender and chech the weight painting and make adjustments where necessary to blend them over the 3 bone limit better.

thanks, I'll look into the weight painting issue. I actually exported this model as SMDs then compiled with Crowbar...pretty sure that doesn't make a difference. If I still can't fix the issue I may package the Blender and SFM files and share them here

[Marco Skoll]

Two big questions: How does it bend in XPS (it's free - even if you don't use it seriously, it's useful to have for porting purposes), and how does it bend in Blender? (You show us the bones, but never demonstrate them bending).

It's important to track down where the problem exists in the chain.

Originally posted by SUPREME:

I actually exported this model as SMDs then compiled with Crowbar...pretty sure that doesn't make a difference.

It does make a minor difference in that StudioMDL (and by extension Crowbar) just automatically picks the three strongest weights on any vertex (while also discarding any of less than 5% if you're not using a hexed compiler) rather than aborting, but in either case the model will only compile if it's been culled down to three bone weights per vertex; it's just which stage it happens at.

[SUPREME]

OK, I got the model to work. The issue wasn't the rig. There were some additional bones within the legs of the Xnalara model when I imported them into Blender, and I had deleted them thinking they were unnecessary. I went back and re-added those bones, and now the model works fine.

The last issue on the model is that a big chunk of her ponytail is transparent when looking at the model from the front (the backside of the ponytail is fine). I'm guessing it has something to do with the normals or maybe i need to play with alpha setting, but I'll leave that for a separate thread.

[SUPREME]

Originally posted by Marco Skoll:

Two big questions: How does it bend in XPS (it's free - even if you don't use it seriously, it's useful to have for porting purposes), and how does it bend in Blender? (You show us the bones, but never demonstrate them bending).

It's important to track down where the problem exists in the chain.

Originally posted by SUPREME:

I actually exported this model as SMDs then compiled with Crowbar...pretty sure that doesn't make a difference.

It does make a minor difference in that StudioMDL (and by extension Crowbar) just automatically picks the three strongest weights on any vertex (while also discarding any of less than 5% if you're not using a hexed compiler) rather than aborting, but in either case the model will only compile if it's been culled down to three bone weights per vertex; it's just which stage it happens at.

thanks, i have XNALara Studio as well. Turns out the issue was my own stupidity in deleting some leg bones in Blender that i thought were "unnecessary" for SFM. Put those bones back in (luckily I save multiple versions of the model as I make changes) and it was 100% good to go.

For the most part. The last issue I'm having is that the ponytail is mostly transparent when looking at the model from the front. Going to fool around with settings and if I can't fix i'll make a separate thread

[Marco Skoll]

Originally posted by SUPREME:

The last issue on the model is that a big chunk of her ponytail is transparent when looking at the model from the front (the backside of the ponytail is fine). I'm guessing it has something to do with the normals or maybe i need to play with alpha setting, but I'll leave that for a separate thread.

That will be a backface culling issue (and a lot of XNA uploaders will mention things about necessary culling settings).

Dependent on the exact application, the options are:

- adding $nocull 1 to the VMTs. Bear in mind that lighting is still calculated from the front face in this case, which means it's really only suitable for the "inside" of things that won't be seen very often (but where you need something to fill the void).
- Duplicating the geometry and flipping the normals on it in order that there are two faces in the same place, facing different directions.

[SUPREME]

Originally posted by Marco Skoll:

Originally posted by SUPREME:

The last issue on the model is that a big chunk of her ponytail is transparent when looking at the model from the front (the backside of the ponytail is fine). I'm guessing it has something to do with the normals or maybe i need to play with alpha setting, but I'll leave that for a separate thread.

- adding $nocull 1 to the VMTs. Bear in mind that lighting is still calculated from the front face in this case, which means it's really only suitable for the "inside" of things that won't be seen very often (but where you need something to fill the void)..

simply adding nocull 1 to VMT did the job. Thanks!

[Marco Skoll]

It's usually okay for hair, but there are some cases where the effect of one-sided lighting just isn't suitable, so the parameter should be used with caution. Like I say, it's generally better thought of for "inside" rather than "back" faces.