M249 - Light Machine Gun

(pasted from Polycount)

I’ve been messing about in Blender, off and on, for the last year and finally decided it was time to take a relatively complex asset through the full pipe without resorting back to Max. The last personal weapon I modeled(tactical shotgun) I used to get the basics of Blender 2.80, but I ended up getting frustrated and went back to my comfort zone in 3ds Max. This time around I had more experience in Blender so it’s 100% done there: modeling/unwrapping/baking/final renders - texturing done in Substance Painter. Target was around 100K on-card verts or triangles. I wanted close up folio shots so I went 4K maps with around 7 texture sheets.

I’ll post a breakdown and more renders over the next few days as I haven’t done one for a while and I’d like to share my thoughts on the process in Blender, particularly, as I baked the bevel shader with no high polys involved at all. I still have some tweaks to do on the texturing(don’t we always :slight_smile: ), but I also set a 3 week time limit so I still have a few days left for final tweaks.

More renders to follow.





Excellent detailing and texturing.

1 Like

Good job! :+1: I wonder why people who already use 3ds max trying to model in blender lately? i’m pretty sure max has more advanced modifier stack than blender. And pure poly-modeling tools are all the same in any dcc. Anyway, looking forward for a breakdown and especially about the bevel shader.


Thanks mate. Appreciate it.

Thanks. I’m finding Blender’s modeling concepts/workflow to be simply faster and more fluid. It’s not about the basic tools. After 15 years in Max it’s starting to feel clunky and dated every time I go back having spent some time in Blender. Also, Hops/Boxcutter/Meshmachine are so good that it’s hard to be without them once you’ve incorporated them to your workflow.

Yes, Max’s modifier stack is far superior, but for authoring game assets Blender has the needed modifiers.

The SAW! Nice!
I’ve humped that thing through many of thorn filled woods and jungles. Saying that, I’ve noticed it’s missing the flip out legs up front… they were vine magnets.

Was glad to make squad leader so I never had to touch that thing again. :grin::joy:

1 Like

Haha! It does look heavy alright. :smiley: Thanks, nice to get feedback from someone who has actually fired one of these. I purposely left the bipod out due to time constraints.

1 Like

Understandable. It looks great and dead on. Brings back memories. :metal:t2:

1 Like

Really great job. The only reason I don’t think it’s a photo is photos don’t usually give that level of detail.

Amazing details :+1::+1::+1::+1:

@Stan.Eshman Thanks man, very kind of you to say.

@AF98 Cheers mate. :+1:

What kind of things you would add to the Blender modifier stack that would improve the gamedev workflow?

To be honest, all the most useful modeling modifiers for game dev assets are already in Blender. The most important one in my workflow is Triangulate. Keeping the asset triangulation non-destructive is the key. One mod that I do miss from Max for this work is the Smooth modifier(non-destructive angle based smoothing groups) The most common mods I always use when modeling are:

Lattice(using an addon)
Simple Deform(bend)
sub-div(if baking high/low)
Bevel(if baking high/low)
Weighted Normals

And that’s it, really, but for things like lattice/bend/circular array I will use addons. Using the native tools for any of these is a real pain compared to Max.

The power of Max’s mod stack becomes more relevant for other kinds of non-destructive modeling workflows, but for pure poly/sub-d modeling Blender equals Max in what the stack has to offer.

What would you add yourself?


Since you asked: I have unfortunately had very little time to do actual modeling so take this with a grain of salt but…

Native VDB objects (and the algorithms to process those) would be nice. As I understand it, the modifier that is likely to be added soon does the mesh <> VDB transformation each time it’s used. Much more performant would be storing VDB and moving it around.

Not sure about modifier, but a way to generate robust semi-functional UV maps that do not have overlaps automatically from meshes would be nice to have.

A little more technical that’s not about modifiers: make the image transfer from and to Python use buffer protocol. Instead of crashing my computer when trying to export glTF with 16k textures, it would handle it without a sweat. https://github.com/amb/blender-texture-tools is where I use CUDA to process image data. It takes 1ms to process and many seconds to move back and forth in Blender. Painful.

Displacement baking. A lot of improvement for baking in general. More samples for bevel shader.

All of these features are a huge amount of work to implement of course, and it’s not even taking into account that it needs to be integrated into the existing solution.


Definitely. A must have.

The gun above was my first time unwrapping in Blender. I bought UVPackmaster immediately as I started and it saved me many hours of work. Having dreaded approaching Blender’s unwrap tools before this project, I am now a solid believer. Still some weird quirks, but overall the simple workflow I developed was lightning fast.

I haven’t dug into this stuff yet. 8K has been the max texture res I’ve exported.

100% on all 3. Couldn’t agree more. Not just the bevel shader samples, though(their is an osl node that can go as high as 512), there are strange baking errors with the shader on certain shapes. There was an in-depth thread about it recently.

Now that UDIMS are implemented, some sort of texture streaming/mipping/atlasing is essential if Blender is to make any inroads into the VFX industry.

I’m a bit behind you, but am making a very similar transition, and it sounds like similar asset production objectives. Even the BoxCutter/HOps/MESHmachine part. For UVs, I bit the bullet at Black Friday and decided to buy a perpetual license for RizomUV, to have a great tool for that function whether modeling in main DCC app or ZBrush. However, I’m finding as I understand more about the native Blender tools that I might get many things done right in Blender. That just wasn’t at all clear when I started putting serious time into Blender a few months ago.

I know you use ZBrush as well. Are you also targeting UE4? I’ll be following along on your work, and echoing what the others have said, this project is turning out really well, and very nicely presented as well.

The “holy grail” objective for me is to settle in on workflows that accomplish the whole high res -> low res -> bake workflow, with generation of the desired, game ready LOD versions, as efficiently as possible.


I bought Rizom a year ago and have the Blender bridge, but wanted to use nothing but Blender’s unwrap tools for this. There were close to 100 parts so it was a good production test.

I’ve been using UE4 for years. The new Epic live link tool has me very excited and is another big boost for Blender as a great game dev hub parogram.

Thank you very much.

Well, from my experience and testing, Blender is more than capable of this goal. The biggest issue at the moment is if you rely heavily on sub-d meshes. The current performance is abysmal in comparison to Max, but it is being worked on.

Thanks for the comment, Tom.

Also, I forgot to mention, Textools and UVpackmaster are essential addons, as far as I’m concerned. :+1: