Importing a Source Filmmaker/Garrysmod Model into Blender

Hello Modelling Community!

I’ve been using Gmod and SFM for posing purposes for around 6-7 years now and I’ve finally decided that it’s probably a good idea to get the hang out of a 3D modelling programm.
I am using Blender and I’m trying to use it as a rendering tool. The basic idea is that I import a Source Model into Blender to pose it and then use all the different tools Blender has to offer to make it look as good as possible. Basically the same as SFM but with Blender.
Correct me if I’m wrong but I’m guessing I need:

–> The model itself
–> All the different textures+bumpmaps and the like via GCFScape
–> The .phys file for the rig, or else posing the model is going to be impossible I guess

So far I can decompile models and get them into Blender via the Source Blender Tools so atleast I got that working by myself.

My question to you is how do I get the textures + bumpmaps and most importantly the rig onto the model for posing purposes?
I have found a video on how to get the textures via GCFScape so I could probably get that working but I don’t know if this covers all the different texture layers like bump maps and the like so I’d also be grateful for indepth inputs on this subject.
I have absolutely no idea how to setup the rig for the model though so any help here is appreciated.

I’m also trying to get the hang of 3D Modelling in general by reverse engineering those models.

I know I am asking for a lot here but I’d be grateful for any help!
Thanks in advance.

  • you only need the .smd, .qc, and .vta to import into Blender ( and vta only if you need flexes. ) Just import the .qc using Blender Source Tools and you’re good to go
  • use VTFEdit to convert from VTF to .tga; these files will load into Blender fine, you’ll just have to link them in the node editor. Look up a tutorial on Cycles and specifically how to use its nodes, which is what you’ll likely want to use.
    -for models with a skeleton, you’ll have to set up IK on your own; a .phys file won’t help.

Unless you are importing a prop, the Source Blender importer has already imported the complete skinned, and rigged model. The .phy is simply defines collision for a game engine environment , its derived FROM the rig, not the other way around. It is unnecessary for your purposes since you are going from source TO Blender.

That being said,

That’s it! You are done! You have imported a Source asset into Blender. You have reached the end of the Source engine-jurisdiction.
From here on out, you are in BLENDER territory, and that is a specialized set of knowledge that is generally not covered in this forum. You may want to hang out with Blender people, not Source-engine people. Once a 3d-asset is in Blender, it is the same as any other 3d-asset, regardless if it came from Source, OBJ, DAE, etc.


The textures are not showing up because Blender has not been provided a **Texture **to render. The Source import has imported the Material, but NOT the Texture. Open the appropriate Texture tab in Blender and apply the desired texture of the model. As of my Blender version at least (2.76b, 2015 build), Valve’s VTF format is not supported. Convert the VTFs into a mainstream format (jpg, bmp, gif, etc.) via the method of your choice, and select the appropriate rendering mode of your choosing.

Thank you for your help guys!

I was able to import a model from SFMLab, with the rig, ready to pose + all the textures and their bumpmaps.
Everything worked just fine, there were just loads of texture groups and assigning them was kinda tedious but it did look really good with the an added PBR Shader.
Now I can experiment and reverse engineer all I want :v:

I’m glad I decided to try this out, it’s really interesting and I’m defenitely sticking with this.
Thank you for your help, really appreciate it!

Can you recommend me some Blender communities beside the obvious ones? Does Facepunch have a 3D Rendering community?

There’s no reason to limit yourself with SFM models. You might as well grab and render the XNAlara version of a model as it has not been already processed (and in some cases, limited) for SFM engine resource limitations (like bones and vertices).

XNAlara provides more variety, and if you use the better importer (downloaded from deviantart), the model is imported already fully textured, so no need to convert VTFs or decompile anything. You have full access to the original bones, because in many cases, some bones were removed to make the model SFM-compliant. (99% chance rig was reduced to make it Gmod-compliant. Don’t bother getting the Gmod-version of a model, unless as a last resort)

The only blender community I personally ended visiting the most was, but that usually was due to searching for the technical answer to something, and not render-based.

To summarize, in my opinion:

  1. Get the XNALara version of the model
  2. If not available, get the SFM version
  3. Very rarely should you end up getting the GMOD version, unless its a custom model made straight-for-gmod (Yes, some people still actually do this)

This really is some indepth input, thanks a ton man, didn’t know the Source Models had to be downgraded so much.
Also it would be really cool if I didn’t have to extract all those textures , I’m defenitely going to check on Deviantart first before downloading anything from SFM or GMOD.
I guess Deviantart is the go to site for XNALara models?

This has been immensely helpful, I’ve been in Blender 24/7 for the past days and it’s been a blast. It’s complicated but really fun, what an awesome tool.

Cycles is great, the rendering is simply beautiful, I’m going to have to look up some lighting tutorials since only very basic stuff carries over from SFM like Three Point Lighting and the like. Only thing I did so far is create the really awesome PBR Shader from Blendergurus Youtube tutorial.

Its because I port and make custom models to gmod as my end-point, and it requires the most processing and resource reduction.

Sure, converting from SFM is easier but I want full control, and I don’t want to deal with a model that has already been processed by someone else. I want the more raw-material.

Its like you want to make something out of play-doh, why use someone else’s used play-doh instead of a fresh can.

By that logic it would be best to skip xnalara entirely and grab whatever model you need directly from either the author or the game.
The less times it’s been run through something the better.

That is true. But some of us do not have the means to purchase the game for the model, or deal with the process of extracting. Or perhaps, we do not have the required hardware or the means to acquire the hardware to extract from certain hardware formats.

I am not disparaging SFM model converters or makers. I certainly, firsthand, understand the efforts and time required to process and convert a model. But for the Original Poster’s purposes, if we were to factor time and convenience, he would be better suited to using XNAlara, the most widespread and popular 3d format for hobbyists (for a good reason). From a porting perspective, using an SFM model is like choosing a “Wizard” install option. There is nothing wrong with that, especially for the less experienced, as some nuanced features have been taken care of. But once a user has amassed more technical experience, he may choose a “Custom” install, allowing him to personally pick and choose the exact features he wants.

XNAlara may be a processed form, from the raw material, but it is guaranteed to be closer to the source material than SFM. In fact in many instances, it has the EXACT original rigging, straight from the source, the only difference being the normalized bone-naming conventions for use with the XPS poser. (So that, for example, if an XPS user wanted to move the head, he knows to look for “head neck upper”, instead of “Bip01Head”, or “jl00191”, or some ambiguous bone name.)

Then you need only ask someone who does have the means to extract them. You’ll often get it.

From my understanding the opposite is true. The xnalara model format does not perserve bone rotations nor explicit vertex normals and only recently obtained support for multiple uv channels. The latter of which is its only advantage over smd.
Perhaps this opinion of your stems from attributing limitations with the mdl format to the smd format. The smd format doesn’t have a limit on bones, weights, nor verts and will keep everything but multiple uv intact. It is a surprisingly good format to transfer between 3d editor with due to import and export plugins existing for nearly every single modelling program on earth.

And then we have fbx. Which is the current standard and for good reason. It kicks the ever loving shit out of both formats.

Source limits bone weight influence to 3, after which any excess is culled to allow compilation. The Original Poster wishes to pose in Blender, and the lack of the bone weight influence allows him to make, say, the clothes “flow more smoothly” for a static render.

By all means, the original poster is free to choose the format of his liking. FBX is a very good format indeed. But I believe he is more likely to find the model he is looking for in XPS. If he has the connections to obtain a more optimized and polished format for his intentions, or if he desires to use the Source Format, that is perfectly fine. I personally, solely, use the Source Format and engine for my hobby, but I like to maintain an open perspective.

Re-read what I posted please. The mdl format is limited to 3 weights per vertex courtesy of the compiler. The smd format is not for whatever reason.
Which doesn’t matter as I was suggesting using neither smd, mdl, nor xnalara.

The issue is, that he sought to obtain SFM models through decompilation. Obtaining SMDs through decompilation will result in the vertices having already gone through the compiler, and hence, culled to the 3 weight influences per vertex. Additionally, decompilation is not a perfect process that perfectly reconstructs a model pre-compilation.

Referring to your previous point, if he has the connections to obtain the model from individuals to avoid the process of decompilation, thats up to him and his personal connections or queries.

You and I, we have the luxury of using SMD as an intermediary format, as we can simply export with the weight cull disabled. For OP, on his own, the only way to replicate a desired SFM model into Blender is through decompilation, giving him the “shafted”, “watered-down”, “un-perfect” version.

As an ending note, sure XNAlara/XPS isn’t perfect. For example, the format cannot carry shapekey data. Either you manipulate the bones or mesh itself to create your expression (this is Blender after all), or just use the SFM version if it has the flexes and you value them above all else. You have infinite options, and there is no right or wrong answer.

(Although you will probably have a interesting time in blender with the eyes if the SFM model had the source engine eyeposing implemented. Good luck with that)