The BRAWL² Tournament Challenge has been announced!
It starts May 12, and ends Oct 17. Let's see what you got!
https://polycount.com/discussion/237047/the-brawl²-tournament
It starts May 12, and ends Oct 17. Let's see what you got!
https://polycount.com/discussion/237047/the-brawl²-tournament
Best Of
Ambrosia Sky enviro art
https://www.artstation.com/artwork/oJvbBw
Yo yo, my studio's game Ambrosia Sky was released yesterday on Nov 10th! I was the enviro artist on the game. I did most of the kits, props, textures, level art and lighting. We had a few other talented artists contract with us for key moments in production.
Yo yo, my studio's game Ambrosia Sky was released yesterday on Nov 10th! I was the enviro artist on the game. I did most of the kits, props, textures, level art and lighting. We had a few other talented artists contract with us for key moments in production.
The game itself is a cleaning simulator, in a brutalist civilization, on the rings of Saturn. You pressure wash away dangerous fungus to clear your way through the game. Everybody who use to live here died because of the fungus. You are sort of an undertaker, learning about the people who use to live there, and then laying them to rest.
Soft Rains art team includes art direction from Adam Volker, concepts from Ysabel Li and Vash Taylor, and tech art/effects by Fiona Jeeva.
Soft Rains art team includes art direction from Adam Volker, concepts from Ysabel Li and Vash Taylor, and tech art/effects by Fiona Jeeva.
I'll be putting up a few more posts containing the art but I didnt want to do it all at once. More incoming!
Ashervisalis
3 ·
Re: Sketchbook: Frank Polygon
Subdivision sketch: it's not the topology.
One question that's asked over and over is: "How can this shape be attached to a curved surface?" At some point, often when just learning how to model, most artists end up asking this question in one way or another.
The short answer is: just match the segments when possible and offset the intersecting geometry so the existing geometry in the curve(s) forms part of the outer support loop around the shape intersection, while also avoiding disrupting the segment spacing or causing unintentional deformation to the curvature and constraining any potential smoothing artifacts to the transitional area between the inner and outer support loops around the shape intersection.
That answer doesn't include any hard and fast "rules" about quads, triangles, poles, etc. because all of that is largely contextual. Depending both on the type of modeling and what the model will be used for. It also doesn't advocate for one singular approach to modeling or a specific topology layout strategy because what works best often depends entirely on the shapes and how they intersect.
Things to watch out for.
A significant problem with trying to formulate a fixed rule set for a universal topology layout strategy (independent of constraints, experience, and goals) is a lot of the commonly parroted one line tidbits of subdivision modeling advice just ends up being absolutist, dogmatic, or reductive. Often in ways that feed into counterproductive habits like analysis paralysis or some form of box ticking perfectionism. Which, to be really honest, is often just an advanced form of procrastination through busy work.
One of the reasons artists can struggle with this is a lot of
art stuff tends to be ambiguous or subjective in a way that's challenging
to learn or just outright not all that fun to push through to see positive results. There's
also a lot of visual problem solving where things don't look right until
they do and progress isn't always linear in that space.
In contrast to that, chasing
technical metrics is a lot easier to gamify and gives a more concrete
sense of progress. Even when the art things aren't actually moving forward. Which is why, more often then not, a lot of what goes into learning and streamlining modeling workflows all comes down to prioritizing what matters.
For most artists, especially beginners, technical things should further art things. Prioritize making art and learning over hitting arbitrary technical metrics that don't directly contribute to the quality and usability of the final assets. That's what matters because that's what ends up in front of the audience.
When it comes to solving modeling and topology challenges: make sample models, test out different approaches, and compare the results. If one approach isn't working then be willing to set it aside and try a different one on a new sample model. Then apply that approach to the working model or merge in the best sample.Below are a few visual examples of what a subdivision modeling based exploration of the initial question could look like.
Example 1: This is where most artists start. Smashing shapes together. Manually cutting in support loops around the shape intersections. Filling in the individual faces and trying to join up all the edges in a way that kinda sorta makes intuitive sense while also following what other models look like.
It seems to subdivide alright. There's a tiny little dent in the highlight where the triangle bridges to the outer support loop near the base of the shape intersection. At first glance it's passable.
Typical one line advice would be to resolve the mesh to all quads or use a different approach to merging the shapes. Bonus points if the advice includes: just use more geometry!
Example 2: The emphasis here is on more geometry, all quads, and only using tools (like inset and loop to circle) to try and redirect the loop flow without having to manually add, remove, or join mesh components. This approach is easy, quick, and checks off all the technical points previously mentioned.
The loop around the shape intersection is clean and visually appealing when the mesh is unsubdivided. It's also been automatically resolved to all quads. When subdivided the mesh appears to be fairly clean but there's some very subtle shadows and highlights radiating out from the base of the shape intersection to the area near the five pointed E poles near the corner vertices.
Typical one line advice for this mesh would be to redo it and match the segments in the two shapes so the poles move closer to the shape intersection.
The loop around the shape intersection is clean and visually appealing when the mesh is unsubdivided. It's also been automatically resolved to all quads. When subdivided the mesh appears to be fairly clean but there's some very subtle shadows and highlights radiating out from the base of the shape intersection to the area near the five pointed E poles near the corner vertices.
Typical one line advice for this mesh would be to redo it and match the segments in the two shapes so the poles move closer to the shape intersection.
Example 3: Here the emphasis is on matching the segments of the two shapes, limiting the number of poles, and using tools to simplify adding support loops. The modeling process is relatively straightforward but there's a choice of whether to join the vertices from the larger shape to the smaller or from the smaller to the larger. Once to two shapes are connected it's easy to select all the edges that define the shapes and add the support loops with a single bevel / chamfer operation.
Again, the mesh is all quads and seems to subdivide cleanly. The outer support loop at the base of the shape intersection was pushed out along the curve and that appears to have minimized the surface deformation where the vertex from the larger cylinder was merged into the vertex on the smaller cylinder.
Typical one line advice here would be to increase the width of the outer support loop and move some of the vertices to minimize the deformation on the outside of the larger cylinder. Alternately, try merging the verts from the smaller cylinder to the ones on the larger cylinder. There's also a six sided star pole that some artists might recommend resolving to a quad grid or some other topology layout.
Comparison of the first three examples: though all of these topology layouts appeared to be passable or at least meet the technical requirements that would typically be offered as one line feedback, it's also important to look at the model from a variety of angles and use a material that highlights any subtle smoothing artifacts or potential surface quality issues that could impact reflections or normal bakes.
Example 1: seemed to subdivide cleanly except for that small smoothing artifact near the base of the shape intersection. When viewed from other angles this does hold up fairly well but there is some very minor surface deformation caused by pinching near the sides of the shape intersection where the edges run out perpendicular to the axis of the larger cylinder. This result is somewhat passable.
Example 2: appeared to be a quick fix since it used all quad geometry and was generated by basic modeling operations like inset, loop to circle, and extrude. There were some very subtle stretching artifacts but they appear to be much more severe when viewed from other angles. So this result isn't all that passable.
Example 3: had some manual cleanup but the modeling workflow could be further streamlined by using a modifier to complete the vertex merge and bevel / chamfer operations. There was some minor surface deformation that was visible with the subdivision wire frame preview enabled and the shading discontinuity from that deformation is still visible but fairly subtle. This result isn't ideal but it's also somewhat passable.
Example 4: is similar to example 2 but uses segment matching aligned to a quad grid to move the poles closer to the base of the shape intersection. So close that it almost appears as if there might not be any poles but there's poles. They just change direction and the one edge runs up off the grid on the larger cylinder and into the segments of the smaller cylinder.
There's some very shallow surface distortion near the base of the shape intersection and a couple of hot spots around the vertex at the center of the poles but, other than those relatively minor issues, it appears to subdivide cleanly.
Typical one line advice for this example would be to add an outer support loop around the base of the shape intersection and move the poles outward to resolve the hot spots in the highlight around the base of the smaller cylinder.
Typical one line advice for this example would be to add an outer support loop around the base of the shape intersection and move the poles outward to resolve the hot spots in the highlight around the base of the smaller cylinder.
Example 5: is a simplification of example 4 and uses less geometry while also routing one side of the outer support loop between the existing edge segments on the larger cylinder. Rerouting the topology flow moves the poles up and away from the base of the shape intersection. Which resolves the hot spot from the previous example and creates a new outer support loop path around the base of the smaller cylinder.
The trade off is adding that support loop between the segments of the larger cylinder introduces some unintended deformation and reduces the width of the highlight at the base of the shape intersection. This appears to be a relatively minor issue and the subdivided mesh seems like it's relatively clean.
Typical one line advice for this mesh would be to simplify the loop flow and reduce the number of poles.
Example 6: is almost identical to example 3 but the loops around the base of the shape intersection were created manually and there's minimal shape deformation on the surface of both the large and small cylinder.
Overall this mesh appears to be clean, with no visible smoothing artifacts, and the edge highlights are fairly consistent. Even though the width of the outer support loop varies.
Comparison of the last three examples: in the end, a lot of the typical one line advice just kind of went in circles. Improving one thing but causing an issue somewhere else. So much so that, in this final comparison, the topology of example 1 and example 3 were more or less correct and the root cause of the issue was unintended surface deformation. Not geometry density, loop flow, triangles, n-gons, or poles.
Example 4: was a definite improvement on example 2 but still had some smoothing artifacts caused by moving the poles too close to the intersecting cylinder and not having an outer support loop path around the base of the shape intersection. In some situations it might be passable but it still leaves a lot to be desired.
Example 5: improved over most of the previous examples. Except that extra edge (used to complete the support loop path around the outside of the shape intersection) disrupts the segment spacing on the larger cylinder and causes a pinching artifact. In that respect example 1 and example three were about equally passable or maybe even slightly better than example 5.
Example 6: is pretty much free of smoothing artifacts and has a relatively consistent surface quality. It's arguably the best but the downsides are the topology routing around the base of the shape intersection has to be created manually and there's a star pole on each side. Neither of which are deal breakers in most scenarios.
Why not just use more geometry?
A couple of reasons. Arbitrarily increasing the amount of geometry naturally constrains potential smoothing artifacts to a smaller and smaller space. Essentially just making the problem so small that it's suddenly not a problem anymore.
That kind of brute force approach to minimizing smoothing artifacts works but often doesn't address the fundamental issue that was causing the problem in the first place. At least not in a way that's applicable to other situations where the geometry density is constrained by other surface features or other parts of the modeling process.
More importantly, blindly throwing more geometry at subdivision modeling problems tends to increase the mesh complexity far beyond what's reasonable and efficient. It's that last part that really degrades the viability of subdivision modeling. Especially when it's compared to more contemporary workflows like boolean re-meshing, bevel shaders, and parametric modeling .
Does topology even matter then?
There's a lot of different answers to that... And by different, I mean mostly wrong.
Not wrong in a philosophical or practical sense. All topology is valid: provided it does what it was supposed to do. It's just some topology works better than others but this isn't even about the topology. At least not directly. Because the short answer to most topology questions is usually one or more of the following answers:
-It depends.
-It doesn't matter.
-It shades correctly, so it's fine*.
*(Outside of very specific technical constraints, especially in terms of most game art, as long as minor smoothing artifacts don't directly affect the visual quality of the normal bake applied to the low poly model then it's often passable. Every discipline, project, studio, etc. has different requirements and if your lead tells you to only use quad geometry with minimal poles then you should probably do that. They can also explain the how and why behind those requirements.)
Subdivision modeling tends to be more approximate than accurate because it relies on the smoothing effect of the algorithm used to subdivide the geometry. Which means things aren't always going to be accurate or perfect and that's ok. As long as the results are visually appealing and consistent.
Does topology even matter then?
There's a lot of different answers to that... And by different, I mean mostly wrong.
Not wrong in a philosophical or practical sense. All topology is valid: provided it does what it was supposed to do. It's just some topology works better than others but this isn't even about the topology. At least not directly. Because the short answer to most topology questions is usually one or more of the following answers:
-It depends.
-It doesn't matter.
-It shades correctly, so it's fine*.
*(Outside of very specific technical constraints, especially in terms of most game art, as long as minor smoothing artifacts don't directly affect the visual quality of the normal bake applied to the low poly model then it's often passable. Every discipline, project, studio, etc. has different requirements and if your lead tells you to only use quad geometry with minimal poles then you should probably do that. They can also explain the how and why behind those requirements.)
Subdivision modeling tends to be more approximate than accurate because it relies on the smoothing effect of the algorithm used to subdivide the geometry. Which means things aren't always going to be accurate or perfect and that's ok. As long as the results are visually appealing and consistent.
Time is a finite resource and keeping subdivision models relatively simple, while also offloading basic modeling tasks to tools and modifiers, helps streamline the modeling process but sometimes the easiest way to be more efficient is to just avoid doing dumb shit in the first place.
What's the benefit of having a mesh that's all quads, with visually appealing loop flow, and minimal poles if the model's proportions are off or the shapes are inaccurate or the surfaces are deformed in weird ways? How much is that technical emphasis worth if it took a lot of extra time to get a result that's just average or not so good? Do any of the previously mentioned technical elements actually move the needle when art fundamentals are lacking?
Sometimes choosing between different topology routing strategies just comes down to making tradeoffs and finding whatever works best with the adjacent constraining features that are already part of the mesh. It's a similar sort of thing with mesh density: the goal is to balance accuracy with editability. Which leads to the concept of letting the shapes define the basic loop paths and refining those loop paths before trying to add a bunch of support loops.
What's the point of all this?
To foster critical thinking. That's kinda it: there's so many sources of information, often of varying quality, and rather than take things at face value, especially if it's just generated words or summaries or random opinions or anything without an actual practical demonstration of the how and why, it's probably worth doing some small scale tests to see what works and what doesn't.
Think of samples and practice models as a way to test something before investing a bunch of time in that particular approach or workflow. Or even just a way to figure out how things work in general.
A lot of the learning required to improve modeling skills comes down to knowledge applied through practice. There's lots of places and ways to get knowledge: reading modeling write-ups, searching for posts in the how do you model this thread, watching other people model, asking questions, getting feedback from other experienced artist, etc.
Engage critically in the search for knowledge and apply that knowledge through practice with active learning. Also, connect with other artists, find a mentor, or work with a team on a project. Whatever suits your learning style. All of those things either provide opertunities to or support practice and skill building.
FrankPolygon
3 ·
Elden Ring Inspired - Marika’s Cathedral of Gold
After her ascension to godhood and before Messmer’s Crusade there was a brief period of peace between Queen Marika and the Hornsent. Built as a celebration of the two cultures, the cathedral thus shares Leyndell’s golden lineage and Enir Ilim’s heavenly spirals.
During the War of the Ancient Dragons however the cathedral was unfortunately destroyed and lost to time.
I’ve been working on and off on this passion project for the past year and I am so excited to finally share it (also relieved it is finally over).
There have been several goals I wanted to achieve and explore with this scene:
-Pay homage to my all time favourite video game Elden Ring (i can finally resume my RL1 playthough now that this is done)
-Push visuals as far as I can with Nanite, Lumen and Megalights and see what I learn from it
-Last but not least, I wanted to see if I’m able to create a scene that can stand shoulder to shoulder with the original game. Let me know if you think I managed to meet this expectation.
The Foliage and the one pile of dirt I used everywhere are Megascans. The Sky is by Velarion and our tarnished warrior is by TalkingDrums.
Everything else was modelled, textured and assembled by me in Unreal Engine 5
The full Artstation post features a bunch of videos I wasn’t able to upload here, please have a look:
https://www.artstation.com/artwork/a0529z
bduck67
5 ·
Re: Do you think polycount a dying platform?
Reflect on Community; the Dutch Oven maneuver requires at least two.
Sharing is caring.
Eric Chadwick
3 ·
Re: Texturing Practice
Okay, Back with another update.
I polished some more finer details ( as if that ever finishes... ) and setup a scene in marmoset i thought would look cool.
I changed some of the camera settings regarding the focus and did a render with those.
I polished some more finer details ( as if that ever finishes... ) and setup a scene in marmoset i thought would look cool.
I changed some of the camera settings regarding the focus and did a render with those.
nyxnim
5 ·
Re: Texture baking features?
Eric please - that's really offensive.Eric Chadwick said:This is why you shouldn't trust whatever the stolen-data-sifting machine spits out as absolute truth. It's straight up lying to you.
the preferred term is "Planet destroying slop pump"
poopipe
3 ·
Re: BRAWL2 [Character] Tracy, pxgeek
Thanks Hawk! Unfortunately I haven’t worked on it since. I basically had to drop everything to take care of a family emergency (still reeling from the loss, tbh).
Hopefully, I’ll be able to come back to it at some point, but for now it’ll stay in my to-do list of personal projects, hehe…
So sorry guys I wasn’t able to contribute more 
pxgeek
3 ·
Re: [WIP] Jedi Realtime Character - Metahumans in UE 5.6
So... I liked my desert environment but wanted a ship background as well. I downloaded a free 'sci-fi' corridor but it wasn't giving the right vibes so I ended up making a super quick ship background last weekend based on some in game screenshots. And look, I'm not an environment artist, I won't pretend to be one, but this does the job for some background shots
I also updated the character textures, pushed the contrast overall and added a lot more variation. Might go back in again for a detail pass if I can be bothered. I also tweaked the face tattoos but now I'm wondering if the forehead design is too small? Also, I updated the lighting in between these 2 images, my b.
I've been looking at poses again along with compositions, lighting etc. and it is HARD
So here's where I'm at so far. Can't decide if I will keep all of these poses but at least I have some choices
I tried out a cloth sim render with wind but it seems weirdly 'overcranked' or something... And the fake hair wind seems to stick... Something to look into...
*EDIT - tweaked this cloth sim and added a teeny tiny shitty animation
Anyway, still need to fix some skinning issues, maybe make some basic ass idle anims to show off the cloth sim like this, and keep thinking about how the hell I'm presenting this
I also updated the character textures, pushed the contrast overall and added a lot more variation. Might go back in again for a detail pass if I can be bothered. I also tweaked the face tattoos but now I'm wondering if the forehead design is too small? Also, I updated the lighting in between these 2 images, my b.
I've been looking at poses again along with compositions, lighting etc. and it is HARD
So here's where I'm at so far. Can't decide if I will keep all of these poses but at least I have some choicesI tried out a cloth sim render with wind but it seems weirdly 'overcranked' or something... And the fake hair wind seems to stick... Something to look into...
*EDIT - tweaked this cloth sim and added a teeny tiny shitty animation
Anyway, still need to fix some skinning issues, maybe make some basic ass idle anims to show off the cloth sim like this, and keep thinking about how the hell I'm presenting this
Re: The Bi-Monthly Environment Art Challenge | September - October (98)
Vastra said:September–October Challenge Submission – Dragon Forge Environment
This concludes my entry for the September–October challenge.
I focused on creating a handcrafted environment and aimed to make the space my own while staying inspired by the provided concept.I chose a dragon theme, a personal favorite, and pushed myself to sculpt, texture, and light the scene in a way that reflects ancient grandeur and mythic character.
This project came with challenges and learning moments along the way, but it was extremely rewarding to work on. I’m very proud of the result and grateful for the opportunity and inspiration. There were moments I had to rethink my approach and learn new tricks, but that’s what made this project fun. I’m very happy with how it turned out!
You may recognize the runes. It is from the game TES V Skyrim, the names of four main dragons: Alduin, Paarthurnax, Odahviing and Durnehviir. The ruby crystal in the center is inspired by more Elder Scrolls lore, the Amulet of Kings.
The red and black banners, a little inspiration from house Targaryen.
I’d love to share the final piece on ArtStation, if allowed.
Thanks for a fantastic challenge!
I'm always happy to improve, so if you spot anything that could be pushed further or refined, feel free to share feedback!
First of all, congratulations on finishing the scene! In my mind that is the biggest challenge of all.
It does feel like a dungeon right out of Skyrim.
A couple of things I feel you could do to improve your scene would be:
- Make a triplanar material/shader, it would solve your tiling issue on the walls without having to modify your UVs
- Try to keep your texel density more even. It is most noticeable in this picture, comparing the resolution of the texture where the dragon is, vs the rest of the pillar.
-Finally for your "hero assets" such as the statues and dragon heads, you could bake a curvature map, to create more interesting variation in the textures. (I couldn't find a tutorial about that but basically you take your high poly, bake only the curvature information in a grayscale map and use that as a mask for material blending in Unreal)
Congratulations again, and keep up the good work!
Pep_mepla
4 ·
Re: The Bi-Monthly Environment Art Challenge | November - December (99)
Cool concepts! I couldn't resist to start the ruins one. How far will I get? 
Just simple modeling, controlling shading via hard edges, don't intend to do any highpoly/ sculpting. Rendering in Unity engine, baked lighting with Bakery. Next, I would apply surfacing, probably using a hand-painted trim atlas. Final result could be a web build with camera controls, would be sweet to have a pawn to walk around the scene.
Well, excited to see what others do with the concepts :-B
edit: Oh wow that yellow looks garish. Should have double-checked the color without f.lux enabled. Ok, updated renders
Just simple modeling, controlling shading via hard edges, don't intend to do any highpoly/ sculpting. Rendering in Unity engine, baked lighting with Bakery. Next, I would apply surfacing, probably using a hand-painted trim atlas. Final result could be a web build with camera controls, would be sweet to have a pawn to walk around the scene.
turntable animation
Trying to keep the damage non-destructive using a boolean modifier. I find it's a bit fiddly, if misaligned cuts don't get capped.
Well, excited to see what others do with the concepts :-B
edit: Oh wow that yellow looks garish. Should have double-checked the color without f.lux enabled. Ok, updated renders
Fabi_G
11 ·