MOD:Movie-Style Holograms for End Game Cutscenes

[DarthParametric 3,783]

View File

Movie-Style Holograms for End Game Cutscenes

---

This mod alters the (frankly terrible) holograms used in the cutscenes prior to boarding the Star Forge at the end of KOTOR to bring them more in to line with the style seen in the movies and TSL. For some unknown reason, Bioware decided to go with a weird texture-based approach that simultaneously looks both horrible and nothing like how Star Wars holograms are typically depicted.

Before:

 

 

After:

 

Changes:

• Admiral Dodonna hologram model completely replaced with a new one based on her regular appearance. No longer has incorrect mirrored torso UVs.
• Dodonna model split into sub-meshes for optimal hierarchical occlusion.
• Vandar model given some mesh tweaks to remove some intersection issues.
• Models now use mesh-based transparency rather than additive texture blending.
• Custom keyframes added to animations to facilitate mesh-based alpha approach and added meshes.
• New animated textures for Vandar and Dodonna created based on their regular appearance textures.
• Minor alteration to one shot during the DS cutscene to prevent Dodonna leaping out of frame while Vandar is speaking.

Known Issues:

• Because meshes do not self-occlude, there are some artefacts during certain shots. For example, seeing the neck through the chin.

Acknowledgements:

• Thanks to @bead-v for KOTORMax and MDLEdit
• Thanks to @ndix UR for MDLops

---

• Submitter
  DarthParametric
• Submitted
  08/18/2018
• Category
  Mods
• K1R Compatible
  Yes

 

[bead-v 251]

So I found the reason why I (and probably no one) can decompile Dodonna's holo supermodel – one of the node offsets is 0 Could I get the ascii for further investigation?

[DarthParametric 3,783]

S_HoloDodonna-mdledit.mdl.7z

[bead-v 251]

How did you even manage to compile it?

[DarthParametric 3,783]

Ah yeah that's right, I forgot the two dummy nodes for the eyelids. MDLOps compiles it just fine. You'd have to ask @ndix UR for the specifics. Given all the hassles with MDLEdit at the time, I had shelved it for work on that mod in favour of MDLOps.

[bead-v 251]

Well, that explains it then!

EDIT:

Actually, @DarthParametric, I have another question. TOR_Hat for example is a dummy in this model. But it has animation nodes with alphakeys. Mdledit will ignore those, because it gets the type of the geometry node, which returns 'dummy', not 'trimesh/skin', so it refuses because dummies don't have alpha controllers. I've never seen this situation in vanilla files. If this is valid, then I may need to change the logic for mdledit.

[DarthParametric 3,783]

A supermodel provides animations. It shouldn't matter what the actual geometry is or if it even exists because you aren't using that. It's just a placeholder for the animations. So I shouldn't have to load pointless geometry into a supermodel that is never going to be rendered in the game. It's just needless bloat. Apparently MDLOps agrees with me.

[bead-v 251]

Yes, normally I'd agree with you, but controllers are stored in the model by number, and some controllers share numbers:

nr 88 = birthrate (emitter) = radius (light)
nr 96 = combinetime (emitter) = shadowradius (light)
nr 100 = drag (emitter) = verticaldisplacement (light) = selfillumcolor (mesh)
nr 132 = p2p_bezier3 (emitter) = alpha (mesh)
nr 140 = randvel (emitter) = multiplier (light)

The tools can only know which controller it is if the type of the node is also present in the model.

We could agree to only turn meshes into dummies, and make mdledit default to mesh (and light?) controllers, but there's no guarantee... The tools can't know if what they're outputting is correct. I find that problematic.

[DarthParametric 3,783]

I only ever added the dummy nodes in the first place to keep MDLEdit from complaining. As evidenced by the lack of eyelid dummies, MDLOps was happy just having the animations, as is the game (assuming MDLOps doesn't add in nodes by itself). This is something for you and @ndix UR to discuss I guess. I'm not particularly fazed by having to adhere to whatever requirements you devise, as long as the tool produces the desired end result.

[Salk 374]

Hey DP!

As I am finally approaching the last segment of my playthrough I had a chance to test this modification of yours which, as usual, I had to manually install because at the same time I am also using danil-ch's Dodonna's Transmission and zbyl2's Restored BIK movies.

Now, things go generally well enough but I have a few issues with the rendering of Dodonna's and Vandar's models. I was wondering if you could take a look at the video showing those issues and perhaps have some idea of why that happens.

The most glaring problems, as you will see, are Vandar's initially keeping his eyes closed when he joins the meeting (and that is not that short a period so it's quite striking) and a series of clipping around Dodonna's face. About the latter, I noticed that in your video that mostly applies to her hand instead.

Here's the video.

Thanks!

[DarthParametric 3,783]

If you are using the old version of danii-ch's mod then that's probably your issue since you nuked the custom supermodels. The newest version of the mod uses module injection and is compatible out of the box. Check your Override folder for the presence of m12aa_c05.dlg/m12aa_c06.dlg. 

@zbyl2 should be shot for releasing a mod that hard overwrites a module and not updating it to use TSLPatcher.

There's no clipping. What you are seeing is a lack of self-occlusion. The description clearly states that transparent meshes do not self-occlude in Odyssey and this results in artefacts.

[Salk 374]

I wasn't using the old version of Dodonna's Transmission.

I suppose the only issue I have is coming from the lack of self-occlusion. I didn't know that's what it's called but I still wonder why it shows differently on my video than in yours.

[DarthParametric 3,783]

Depends on how you merged in the zbyl2 mod, since you said you "manually" installed it.

Edit: Here, try extracting this into your Override and see what happens.

Alt_Stunt_Models.7z

[Salk 374]

Thanks, DP!

I give it a try now.

About the mod mix I personally have, I don't believe I caused any problem by manually installing your mod and danil-ch's. Following the TSL Patcher instructions to do manual edits is not really rocket science. And zbyl2's mod is just a line of code playing a movie. I added an Entry to the .dlg file to do just that first and then the rest follows.

Alright, the new Vandar model does solve the issue with his eyes being initially shut. Well done!

Unfortunately, the new Dodonna model doesn't make any difference.

I am attaching a few screenshots here to show what her face and hand looks like: 

[DarthParametric 3,783]

35 minutes ago, Salk said:

Alright, the new Vandar model does solve the issue with his eyes being initially shut. Well done!

That's not a good thing. It means you screwed up. Guess it is rocket science after all.

36 minutes ago, Salk said:

Unfortunately, the new Dodonna model doesn't make any difference.

Of course not. Again, this is an inherent engine issue. It will never be fixable.

[Salk 374]

I'm not sure how I have screwed anything up because the Vandar model is not replacing anything in my Override folder.

You are also saying that it was expected that the Dodonna model would not solve anything so what was the point of testing it?

[DarthParametric 3,783]

c_holovandar and c_holododonna are the vanilla stunt models for those two scenes. My mod doesn't use the vanilla stunt models, it uses custom ones. Your install was broken because you didn't install the mod properly and thus Vandar wasn't using the custom stunt model, he was using the vanilla one. What you downloaded was my custom stunt models recompiled to use the vanilla stunt model names. Dodonna looked fine, but I figured if you screwed up one then chances are you'd screwed up both, so I included Dodonna's in there as well just in case.

[Salk 374]

Alright, I have screwed things up.

It seems that K-GFF didn't save my manual changes to the Stunt List I am sure I made. 

I do many such manual edits and it can happen to make a mistake sometimes.

I guess that the self-occlusion issues are worse in my video because of the different camera angles used (they come from danil_ch's revised camera).