Can you show what you mean? I don't see any difference
Models by Jorzi (AR)
Re: Models by Jorzi (AR)
MaNGusT wrote: ↑05 Jan 2020, 01:44In wz it works perfectly too. I don't know why they decided to wait 3.4.0 for such update. Non of us are familiar with vulkan code so it will take ages to rewrite it from ogl. I hope they will change their decision and release it for 3.3.* and then we could switch our workflow for a fastest way.
The goal is to support multiple graphics backends, for as wide support / compatibility as possible.Jorzi wrote: ↑05 Jan 2020, 15:53 Personally I don't see the vulkan API as that necessary for modernizing the graphics. I thought the main reason for switching to Vulkan was that we have developers who are enthusiastic about the new api and that it would make development easier? The previous decision to switch to opengl ES seemed more logical to me, since we want compatibility with as much hardware as possible.
So we'll actually be supporting:
- OpenGL
- OpenGL 3.0+ Core Profile (default), OpenGL 2.1 Compatibility Profile
- OpenGL ES 3.0 / 2.0
- Vulkan 1.0+
- DirectX (via LibANGLE) on Windows (OpenGLES -> DirectX)
- Metal (via MoltenVK) on macOS (Vulkan -> Metal)
The good news is that the multi api PR now includes the optional per-vertex normals support! (With huge thanks to MaNGusT who helped track down some issues impacting the Vulkan backend.)MaNGusT wrote: ↑05 Jan 2020, 19:45as far as I understand, they want multi api support for wz but it will still require some code changes to optional normals directive. Also, I don't know who will fix that since I-Nod is not a graphical engine developer, it took some time to google and write necessary code for right "tangents". Hundreds of holywars about how it should look like, inverted Y channel that can't fix and other good stuff
My thoughts are that we need this normals directive today not tomorrow and I absolutely don't care on what api it will work. I just want to create models with standard technique(without limitations) and put it to mod.
https://github.com/Warzone2100/warzone2100/pull/489
Re: Models by Jorzi (AR)
Thanks for working on these things pastdue & MaNGusT
After some very slight modifications to the fragment shader, I was able to render both my old models, the new tangent space models and the vanilla models at the same time.
Edit: Attaching my edited shaders for reference
After some very slight modifications to the fragment shader, I was able to render both my old models, the new tangent space models and the vanilla models at the same time.
Edit: Attaching my edited shaders for reference
- Attachments
-
- tcmask.frag
- (4.99 KiB) Downloaded 318 times
-
- tcmask.vert
- (2.17 KiB) Downloaded 322 times
-insert deep philosophical statement here-
Re: Models by Jorzi (AR)
I still don't understand why do you prefer to change the strenght and color of lights in shaders instead of changing textures?
Also, don't we focus on recreating vanilla models into new one?
My thoughts are that shaders have to be universal to any artists, not only us.
Also, don't we focus on recreating vanilla models into new one?
My thoughts are that shaders have to be universal to any artists, not only us.
Last edited by MaNGusT on 02 Mar 2020, 04:04, edited 4 times in total.
Re: Models by Jorzi (AR)
It's best if we can stabilize on a pipeline with a single set of WMIT + WZ shaders that yields predictable, consistent results for everyone. (Ideally, no shader modifications should be needed by the ArtRev mod, or anyone wanting to do art / texture work.)
If additional parameters are needed in the PIE files (for example) - that's something I'm happy to explore.
Re: Models by Jorzi (AR)
the list is small but hard
1) allow event-driven animation for all types of models.
2) a new type of connectors should be added. It's called Pivot Point. One per pie level. If specified, it should replace the calculated center of pie level. The main reason to use it is to fix aiming animation with new models. (when a barrel intersect its turret model when an enemy is too high/low)
3) need a way to disable terrible default recoil animation that uses scaling and a way to enable a new recoil animation without scaling in selected pie level of gn* models.
4) a simple SSAO would be perfect. It could add a lot of realism and "atmosphere" to plastic world of wz.
5) return the FOG
Btw, when optional normals will be released to 3.4.0, I-Nod will help me with normals for terrain.
Re: Models by Jorzi (AR)
Oops I left my custom values in the shader code. I like my light to be slightly yellow and my shadows blue. That was actually not my point. Personally I think Ambient 0.5, Diffuse 1.0 and Specular 1.0 are good enough for now, although in the future, once the terrain is also lit the same way, I think it would be a good idea to unlock it for the map makers to play with (i.e. make it map specific).
This.1) allow event-driven animation for all types of models.
2) a new type of connectors should be added. It's called Pivot Point. One per pie level. If specified, it should replace the calculated center of pie level. The main reason to use it is to fix aiming animation with new models. (when a barrel intersect its turret model when an enemy is too high/low)
3) need a way to disable terrible default recoil animation that uses scaling and a way to enable a new recoil animation without scaling in selected pie level of gn* models.
4) a simple SSAO would be perfect. It could add a lot of realism and "atmosphere" to plastic world of wz.
5) return the FOG
-insert deep philosophical statement here-
Re: Models by Jorzi (AR)
I finally finished remaking all the 10 different hut designs and exporting them.
BTW I noticed that WMIT sometimes loads normals incorrectly for pie files with normal data.
Also I tried replacing my custom light values with the default ones and it looked fine so I won't be messing with those anymore
BTW I noticed that WMIT sometimes loads normals incorrectly for pie files with normal data.
Also I tried replacing my custom light values with the default ones and it looked fine so I won't be messing with those anymore
-insert deep philosophical statement here-
Re: Models by Jorzi (AR)
I see you added some changes to shaders. I have a few questions
1) You did -Y to object-space maps to not do invert to blue channel in dozens of textures? Yeah, we probably need a batch converter
2) What is the point to use a normal map without a specular map?
btw, good big update for arizona. hope someday we will finish first campaign.
There are still few models left, afair:
-new textures for python body
-mortar emplacement
-Broken crane
-broken cars and wreckages (I've already started to work on it)
-few big neutral buildings (I've already started to work on it)
-babas buildings
1) You did -Y to object-space maps to not do invert to blue channel in dozens of textures? Yeah, we probably need a batch converter
2) What is the point to use a normal map without a specular map?
btw, good big update for arizona. hope someday we will finish first campaign.
There are still few models left, afair:
-new textures for python body
-mortar emplacement
-Broken crane
-broken cars and wreckages (I've already started to work on it)
-few big neutral buildings (I've already started to work on it)
-babas buildings
Last edited by MaNGusT on 08 Mar 2020, 14:02, edited 1 time in total.
Re: Models by Jorzi (AR)
Hi
1) Yes, we probably need a batch converter All I can say is, the shading was correct this way
2) We have a specular map, but for legacy reasons, it's in the alpha channel of the normal map. This correction could also be removed with a batch converter which extracts the specular map.
One addition to the list would be the scav soldiers, which could be skeleton animated and use tangent space maps now that we have the tech for it
1) Yes, we probably need a batch converter All I can say is, the shading was correct this way
2) We have a specular map, but for legacy reasons, it's in the alpha channel of the normal map. This correction could also be removed with a batch converter which extracts the specular map.
One addition to the list would be the scav soldiers, which could be skeleton animated and use tangent space maps now that we have the tech for it
-insert deep philosophical statement here-
Re: Models by Jorzi (AR)
but we don't have skeleton(joint) animation support. Joint animation could be used for good build animation or advanced fire,reload, recoil animations, or cyborgs movement.. Ah, dreams.
On the other hand, almost all above could be repalced by object animation with pivot points
Re: Models by Jorzi (AR)
My blender export script automatically takes joint-based skeleton animation with constraints, inverse kinematics, interpolation and all the good stuff and converts it into per-frame absolute transformations. The only limitation is that there can be no deformation or multi-bone transformation, except for scaling. This means that each limb has to be its own part, like a pinocchio doll, but at our scales it looks just fine.
Since the firing, movement and death animations are actually not hardcoded for the scavs, this should be easier than replacing the recoil animations.
Edit: I think I just described exactly what you meant by thisOn the other hand, almost all above could be repalced by object animation with pivot points
Since the firing, movement and death animations are actually not hardcoded for the scavs, this should be easier than replacing the recoil animations.
-insert deep philosophical statement here-
Re: Models by Jorzi (AR)
Yeah I think it should be possible
-insert deep philosophical statement here-