Originally posted by Pte Jack:

Nope, I'm not seeing anything out of the ordinary here, unless your Idle sequence is using a SMD that was exported before you made any changes and didn't re-export after them.

Its funny you mention the Idle sequence part. It was something which I had real trouble getting my head around. So I just downloaded the zip file from the Valve Developer Community Blender tutorial I was following, and just copied and pasted every time I made a new model.

https://developer.valvesoftware.com/wiki/Modeling_props_with_Blender

http://www.bagthorpe.org/bob/cofrdrbob/blender/

Edit: The Idle folder part

$modelname "props_lowpoly\RK_LowPoly_Stone_Fireplace_01a.mdl"
$cdmaterials "models\props_lowpoly"

$texturegroup "skinfamilies"
{
{ "RK_LowPoly_Stone_01" }
}

$scale 100.0
$surfaceprop "concrete"

$StaticProp

$bodyGroup "Body" {
studio "RK_LowPoly_Stone_Fireplace_01a.smd"
}

$Sequence "idle" {
"anims\idle.smd"
fadein 0.2
fadeout 0.2
fps 1
}

$CollisionModel "RK_LowPoly_Stone_Fireplace_01a_phys.smd" {
$mass 400
$inertia 1
$damping 0
$rotdamping 0
$concave

}

Still giving me the same results for both of those suggestions.

I'll post the Crowbar.exe compile log just in case there might be a problem there.

Compile log

Compiling with Crowbar 0.34.0.0: "**\****\******\*******\My Materials & Models\rks_materials_models\low-poly\Blender\models\props_lowpoly\RK_LowPoly_Stone_Fireplace_01a\RK_LowPoly_Stone_Fireplace_01a.qc" ...

Compiling ".\RK_LowPoly_Stone_Fireplace_01a.qc" ...
qdir: "c:\users\rupert\desktop\my materials & models\rks_materials_models\low-poly\blender\models\props_lowpoly\rk_lowpoly_stone_fireplace_01a\"
gamedir: "C:\Program Files (x86)\Steam\steamapps\common\Counter-Strike Global Offensive\csgo\"
g_path: "RK_LowPoly_Stone_Fireplace_01a.qc"
Building binary model files...
Working on "RK_LowPoly_Stone_Fireplace_01a.qc"
SMD MODEL RK_LowPoly_Stone_Fireplace_01a.smd
SMD MODEL anims\idle.smd
SMD MODEL RK_LowPoly_Stone_Fireplace_01a_phys.smd
Processing LOD for material: RK_LowPoly_Stone_01
Model has 6 convex sub-parts
Collision model completed.
---------------------
writing C:\Program Files (x86)\Steam\steamapps\common\Counter-Strike Global Offensive\csgo\models/props_lowpoly\RK_LowPoly_Stone_Fireplace_01a.mdl:
Loading preexisting model: C:\Program Files (x86)\Steam\steamapps\common\Counter-Strike Global Offensive\csgo\models/props_lowpoly\RK_LowPoly_Stone_Fireplace_01a.mdl
Found 1 preexisting sequences.
bones 964 bytes (1)
animation x y ips angle
animations 112 bytes (1 anims) (1 frames) [0:01]
Building sequence index remapping table...
No re-ordering required.
sequences 220 bytes (1 seq)
hdr@0054EFE4=79C30200
ik/pose 164 bytes
eyeballs 0 bytes (0 eyeballs)
flexes 0 bytes (0 flexes)
hdr@0054EFE4=79C30200
textures 72 bytes
keyvalues 212 bytes
hdr@0054EFDC=79C30398
hdr@0054EFE4=79C30200
bone transforms 0 bytes
bone flex driver 0 bytes
bodygroup presets 0 bytes
Collision model volume 22896.00 in^3
collision 0 bytes
total 2000
---------------------
writing C:\Program Files (x86)\Steam\steamapps\common\Counter-Strike Global Offensive\csgo\models/props_lowpoly\RK_LowPoly_Stone_Fireplace_01a.vvd:
vertices 11328 bytes (236 vertices)
tangents 3776 bytes (236 vertices)
total 15168 bytes
---------------------
Generating optimized mesh "C:\Program Files (x86)\Steam\steamapps\common\Counter-Strike Global Offensive\csgo\models/props_lowpoly\RK_LowPoly_Stone_Fireplace_01a.dx90.vtx":
body parts: 8 bytes
models: 20 bytes
model LODs: 12 bytes
meshes: 9 bytes
strip groups: 33 bytes
strips: 35 bytes
verts: 2124 bytes
indices: 984 bytes
bone changes: 16 bytes
everything: 3265 bytes

Completed "RK_LowPoly_Stone_Fireplace_01a.qc"
... Compiling ".\RK_LowPoly_Stone_Fireplace_01a.qc" finished. Check above for any errors.

... Compiling with Crowbar 0.34.0.0: "**\****\******\*******\My Materials & Models\rks_materials_models\low-poly\Blender\models\props_lowpoly\RK_LowPoly_Stone_Fireplace_01a\RK_LowPoly_Stone_Fireplace_01a.qc" finished.

Okay, gave that a go. Sadly, it has made no difference.

I have no clue what is causing this issue. I am pretty sure my technique for creating the model is fine (and to be honest, it isn't a particularly complicated model). I do not know whether this could be a .VMT issue, or whether is it a problem with Blender?

I was using the Triangulate module to turn my Quads into Tris(s) before exporting. I found that it minimised the potential for BadVertexData. So I was already tris(ing) it :)

Very strangely, it would seem that the process of the Source Blender exporter, where it converts the model to a .SMD, automatically marks sharps. I imported the .SMD back up into Blender and oddly enough, it seems to have marked a load of new sharps. It seems to be doing it randomly along flat faces.

Left Original | Right Imported
https://imgur.com/QDuDCkj

Hmm, it didn't make any difference.

To be honest, there were no strange vertices out of position like the image you posted. Everything look pretty normal and how it should look in the model viewer in Blender. The normals are fine and facing the right way. There were no doubled vertices or anything like that.

I just don't know why the model is being lit this way.

Closer view of the one of the problem areas
https://imgur.com/ekUdyPQ

So it seems that every quad face gets split into two tris(s) and from the strange lighting after the compile, it seems that only one tris gets lit while the other one stays in shadow. I have no idea why it keeps doing this!!?!?

Originally posted by Pte Jack:

Only a suggestion...
Check inside your model for hidden faces. You may have joined a couple of opposing verts creating faces inside the mesh..

Edit, also check for non manifold vertexes, you may have disconnected doubles.

When using the edge split modifier, you are splitting the two vertices at 90°. So you would effectively have non manifold vertices at every sharp. I don't know of any other way of creating sharp edges on a model.

When disabling per-vertex lighting, it makes no difference. Also have the prop as a non-static prop made no difference either.

I have been creating some low poly assets for a map I am working on. When I compile the map using the full compile to test how the lighting looks, the shadows on my prop_static's look terrible. The props have sharps marked and I have used the edge split modifier to get sharp edges.

What it looks like in game
https://imgur.com/phNfOCw

https://imgur.com/ca7MqF6

Co-locatedVerts view
https://imgur.com/DusQIVG

These strange shadows seem to appear where two tris(s) join together on a flat face. It also seems to glow in strange places around my model too...

Anyone else had/seen this problem before?

Originally posted by shotgunefx:

While more vertices will improve shading, It still looks wrong to me. Just from bounced light, they shouldn't be that black. Are your normals good? I've had instances where a normal was reversed and the result was black faces. Also if those vertices are embedded in the jam , they will be self shaded, and not get light. Try disabling self shadowing on the prop.

It might be hard to see from the photos, but the sliding sash window is split into 3 separate props. The frame round the outside is a prop_static, then the two inner parts which hold the glass are prop_physics. (The glass is a func_breakable, so I dont need to worry about that).

I've checked the normals on all of the models in Blender, and they are all fine. (I normally model with Backface Culling ticked on).

Is there no other way that Source simulates shadows on props other than by vertex? Nothing that potentially could be done via the texture of the model. (Kinda defeats the whole point of low poly if the simple window frame is like 200+ vertices.)

I'm trying to spend a lot of the time on how light and shadows effect the models, becuase thats where a lot of the detail comes from with low poly.