Jeah rly sad...but if you like you can try on your own again if your exporting is finished...maybe i did something wrong
I'm at a jazz festival now but will check later. Export should be done.
I gues we have a big problem...
I cant manage to replace a whole Proxymesh with the puzzled parts of the main meshes because the textures of the main meshes are partly in the mesh itself and another part is loaded with the commands in xml file-.-
Problem is that every puzzlepart has the same material names-.- so House1upper has material 10,11,12 and House1lower has also 10,11,12 so you cant use the same numbers neither in blender nor in the xml files...
was a pleasure to work with you :\
I cant manage to replace a whole Proxymesh with the puzzled parts of the main meshes because the textures of the main meshes are partly in the mesh itself and another part is loaded with the commands in xml file-.-
Problem is that every puzzlepart has the same material names-.- so House1upper has material 10,11,12 and House1lower has also 10,11,12 so you cant use the same numbers neither in blender nor in the xml files...
was a pleasure to work with you :\
yep, that's about what I figured. if we want to merge them we'll need to do it the hard way - with retextures.
I'm still going to try and get the proxymesh=true thing working. that would be the easiest to edit. failing that, I'll just make them high quality up to the proxy mesh.
After all my adventures, i'm not convinced about the grass. Random data
1) Whatever ring multiplier setting you have, hop on roach and run and at least on my machine you're always guaranteed to see a grass ring. Similarly hop off and it goes to infinity again.
2) We all found our favourite number to make our test saves go to infinity.. but.. ive seen much smaller values achieve the same infinity... in different areas.. As in the observed behavior didn't seem like it was in any liner way effecting that quality ring size.. i was lowering values to the point where i would have certainly got a ring in one scene, but gave infinity in another... Could be zone, could be density.. skellege was a problem...
Just randomly for your time, perhaps this isn't the concern in this thread, but meshrenderingdistancescale is relatively harmless on performance (perhaps except on skellege islands).. TextureStreamingHeadsDistanceLimit is what completely kills it, on my setup at least.
As far as the character popin goes.. there's something hard happening there.. we've demonstratably increased the distance low quality meshes appear, but there's a certain class of them that the engine is commanding to popin. Go to the town at winespring grange, its guaranteed there. Great for experiments.. Its certainly not the streamer.. you can go there, up to say 4 meters away.. and be able to mouse target the invisible npc's.. you can do this for days. walk away in the scene, come back, jump sideways and be silly. The only way to see them doesn't matter what is to get into the 1-2 meter teleport class npc distance range. There's something special going on here. ANd yeah, setting the background npc setting to lower than the low value via the ini file does not fix this.
So with the grass ring, the reason it appears visible on horse is because you're up high and moving faster. The grass ring is set at the same distance no matter what's going on. The reason it looks like it goes to "infinity" is because it's hiding in the tree lines, or is far enough out the detail is lost in pixels.
MeshRenderingDistanceScale is broken. It does nothing. It's hard locked at ".5" and doesn't even do anything then, because the values in mesh xmls dictate popin distances, not the scalar. If the MeshRenderingDistanceScale value meant anything, we'd see assets rendering at half the distance they normally render at.
the character TEXTURE popin can be solved with streamer edits. character MESH popin is locked into the xml files, just like all other meshes. I'll be looking into extending these values in my mod, forcing continuous LOD0 (high quality mesh) and no LOD's on an optional character file. I'll just need to re-export all of them because I had to delete that folder to make space for the architecture folder (which arguably needs it more)
Wasn't The Witcher 3 possibly getting DX12? If so, this type of thing should really have an easier time then, right? You could ramp things up a bit.
Nobody from CDPR has ever said TW3 will get DX12, it's always just been a lot of OTHER people saying it, so it shouldn't be a consideration here.
That's right and I am sure I read somewhere, they also said they did not think DX12 would lead to performance improvements on XB1, so I assume it will be the same for PC but PC has a wider hardware profile. Can't remember where I read that, will post back with the link if I find it. I'm not sure how much more the game can be optimised though. I am running an Alienware M17xR4 with an AMD 7970m and the game looks beautiful and plays beautiful.
Last edited:
They never said that it will get DX12 support but what lead engine programmer Balázs Török DID say is that
http://wccftech.com/witcher-3-cdpr-dx12-availabletemporal-aa-consoles/
Yes, which confirms my earlier post. They've never said it will get DX12.
I'm happy to be wrong about this, but in my experiences it does. To test it winespring grange is a good example, but also the approaching the villiage rannvaig from the 2 huts south east of it is a pretty consistent test.
Setting the value to default, as you approach the village, there's a red guard right on the path, that noticably pops into view at quite a short distance. Setting meshrendering to 2 or more produces a noticably different result.. instead of an immediate teleportation of a high quality model, what happens on my setup at least is from a far distance you can see the model, im guessing at a low quality mesh, in a very odd pose with both arms stuck out in a triangle. You can see this in winespring grange also with some of the npcs. If you lower the value back down, you don't get the triangle arms npc at a greater distance.. just the popin.
There's another npc that's sitting on the seat near the alchemist there which 100% does the 2 meter popin at default values, but with the higher mesh render value is always there as you get there.
The lower mesh values do stick out when you notice them.. but having the npc not teleport is enough to suspend disbelief. The setting doesn't remove popin 100%.. just puts it back to a level where its okay. Subjective assessment for playing, for discussing mod features im sure something more absolute should is a better intent.
I promise you that value does nothing. I've spent the past two days trying it to be damn sure. I have run identical paths through novigrad, oxenfurt, velen, skellige, and some caves.
---------- Updated at 04:18 PM ----------
OKAY guys, I can't get xml files to do ANYTHING when I re-import files. I don't know why at ALL.
what static mesh has incredibly obvious LOD popping so I can run tests on it (that ISN'T switching to proxymeshes.)
right now I'm exporting all the proxy meshes, because apparently I forgot to do that.
---------- Updated at 04:22 PM ----------
@marvelmaster - in those youtube videos you posted, did you make an edits to the xml of the main mesh, or was it only to the xml of the proxy mesh? I ask because I don't see an LOD popping before the proxy mesh replaces it.
Last edited:
In the video i guess i did not mod the main mesh...transitions seem very good...but u CAN edit them to get even closer meshhidedistance lol^^
And i realy dont know why the opposite direction does not work
...We may have missed smth
Last edited:
I'll start work with proxy meshes as soon as I get them exported. I might have a way to make a single mesh out of all the "puzzle meshes" as you put it. I think I can import all of them into maya and then export them as a single mesh.In the video i guess i did not mod the main mesh...transitions seem very good...but u CAN edit them to get even closer meshhidedistance lol^^
And i realy dont know why the opposite direction does not work
We may have missed smth
I think I realized what my problem was with xml files. Because I wasn't editing the proxy mesh xml as well, it was popping in way too early. I've fixed that and I'm building the mood now to test all this on the siege tower in skellige.
Last edited:
i managed to inserte the Proxymesh as whole model BUT the textures are the problem :\...the only problem :\
Last edited:
It would be some serious work but I think that we could use Maya to stitch the textures together. I've been unable to load all the meshes in at once though because each mesh contains LoD meshes that are named the same, and Maya doesn't like that. I'm focused on xml edits, though, because those are easier and I can write a script to edit all of them at once once I figure out what needs tweaking.
What has changed in level design precisely? I thought it only got better and bigger with time but I might be wrong :/
So if you got knowledge about modeling and get textures on these you just need to change the texture of of the proxymodel...its 1 file(i guess)
I cant export xbm files so i never found out what kind of texture is located in xxx_proxy.xbm :\
Problem is... Proxymesh=1 texture(i guess) Mainmesh= 30 textureparts-.-(materials? have no idea about modeling so dont realy know the difference^^)
Btw. i have NO idea about scripting but does "import final function ForceGraphicalLOD( lodLevel : int ); " do anything in world.ws file?
I cant export xbm files so i never found out what kind of texture is located in xxx_proxy.xbm :\
Problem is... Proxymesh=1 texture(i guess) Mainmesh= 30 textureparts-.-(materials? have no idea about modeling so dont realy know the difference^^)
Btw. i have NO idea about scripting but does "import final function ForceGraphicalLOD( lodLevel : int ); " do anything in world.ws file?
Last edited: