Best Of

Re: Early-Late 2000s CGI/Game textures

Hi there,

I'd say you are thinking about this backwards. You are assuming things about the past based on what you know of game art and 3D software of today, which I suppose is an intuitive thing to do ... but this leads you to imagine things about previous tech and workflows that just don't apply at all.

- First off, metalness and roughness are just greyscale masks. They can of course be be baked down from complex materials (like the materials one may create in Substance) but that's in no way mandatory. These greyscale masks can be created any way one wants, and because of that the only thing needed from a texture baking tool is the ability to transfer a simple texture map (regardless of what it may be called) from one object to another. As a matter of fact, for some stylized art styles painting the needed masks by hand (either projected on the model or directly on the 2D texture) can still be much faster than attempting to create a complex Substance material and baking it down.

- Secondly, just because an older game looks great doesn't mean that it uses bells and whistles analogous to today standards. Games from the PS2/PS360 era managed to look great with very simple rendering tech. It is perfectly possible to make a character look fantastic with just a well painted diffuse map and a simple mask driving specular on/off. As a matter of fact this simplicity is IMHO specifically what made some older games visually striking, because it favors a strong differentiation of materials (shiny/not shiny).

Here's a rip from The Boss from MGS3, loaded into XNAlara with a very simple material setup applied. The skin of her face looks matte, and her outfit looks shiny. But this doesn't involve any baked metalness or roughness (or any handpainted metalness or roughness either for that matter) - just a low specular value applied to the face and a high specular value applied to the outfit. That's it ! It doesn't involve any painted masks. And the game as whole doesn't use anything more complex than that on anything really. It's enough to convey a whole world of rich locations and characters.

You can see this effect (or rather, this design choice) very clearly in game during the first Sokolov cutscene. His black leather coat has strongly painted highlights in the diffuse texture to take care of most of the look, and it is enhanced by a bit of dynamic spec applied to the whole garment. It's as bruteforce as it gets but it works.

I would highly recommend you to look into actual asset rips from any game you are interested in to see the texture sets involved. Or even better, attempt to mod your own models into PC games from a few gens ago with a strong, still active community (like for instance games that are being community-enhanced with highres models and texture packs). There's a ton to learn from poking around.

And it will also shed some light onto why the way we are building game assets now is incredibly resource-intensive, compared to the barebones (but efficient) processes of 20 years ago.

pior

4 ·

Timeless Epoch (Procedurally Generated)

Artstation: https://www.artstation.com/muqaddasaman

Hey Polycounters! Just wrapped up my final MA project — Timeless Epoch — and I’m super excited to share it here.

1. Initial Inspiration

The idea for Timeless Epoch began with my fascination for base-building mechanics in V Rising. I was especially intrigued by how rooftops were generated automatically whenever an enclosed structure was detected. That simple, elegant solution sparked a question that became the backbone of this project:

“Could an automated system empower artists to generate game-ready structures while still matching the quality of handcrafted environments?”

This question aligned perfectly with my passion for environment art and became the challenge I set for my Master’s final project. My visual benchmark throughout was Oscar’s House from Baldur’s Gate 3, created by Jef Simons and the team at Larian Studios.

2. Creating the PCG System

At the technical core of Timeless Epoch lies a Procedural Content Generation (PCG) system built entirely in Unreal Engine 5. The system uses:

Spline Points → to define the structure’s footprint and height.
Data Assets → a library of modular components such as walls, pillars, windows, and roofs.
Blueprint Graphs → which control how assets are placed, offset, and replaced dynamically.

I expanded this base system with custom logic for:

Trigger boxes with tags → allowing unique props (like windows or doors) to automatically replace spline points.
Corner detection → solving one of the hardest challenges of modular PCG generation.
Custom floor controls → enabling unique designs for ground and top floors

The result is a spline-based system where entire medieval structures can be generated instantly, with artists free to customize parameters like height, wall type, or decorative elements.

3. Creating the Modular Asset Kit

While the PCG system provides the brains, the modular asset kit serves as the heart of the project.

Over 20 unique modular pieces and props were hand-modeled in Blender and textured in Substance Painter/Designer.

Materials were crafted to match the visual quality of Baldur’s Gate 3 and A Plague Tale: Requiem.

Image: https://us.v-cdn.net/5021068/uploads/editor/ht/jazxantkpwcy.png

To compensate for the lack of vertex painting in the PCG workflow, I created alternate wall variations (plastered, mossy, damaged) divided across ground, middle, and top floors.

Image: https://us.v-cdn.net/5021068/uploads/editor/iy/hrhawhpiclg0.png

Decorative props such as torches, chandeliers, pottery, and a custom-designed entrance door added richness and variety to the generated buildings.

The goal was always to build assets at a quality where, even though procedurally generated, the structures would still stand shoulder-to-shoulder with handcrafted works.

4. Final

After months of iteration, the final environment was set as a villa-inspired layout, surrounded by foliage and stone walls (using Megascans for environment props to save time). The focus was deliberately on showcasing the PCG system and modular kit rather than environmental clutter.

The project concluded with a two-minute promotional video highlighting both the artistic and technical aspects of the system. This milestone marks the culmination of my MA Games Art and Design journey, where I transitioned from an artist experimenting with ideas to someone building full-fledged procedural pipelines for game-ready environments.

Most importantly, the project proves a key point:
Procedural generation doesn’t replace artistry — it amplifies it. By automating repetitive structural building, artists can spend more time where it matters: pushing the detail, storytelling, and beauty of their modular assets.

Here is a link to complete project breakdown on my Artstation: https://www.artstation.com/muqaddasaman

Link to a video breakdown of the project: VideoBreakdown_TimelessEpoch

MFThree

1 ·

Re: Stuck: My Topology Journey Has Hit a Wall

FrankPolygon said:

@Gikkio Although there is some overlap in the basic art, observation, and problem solving skills required: concepting and subdivision modeling are different disciplines. While it may be possible to learn both at the same time it tends to be easier to focus on mastering each discipline separately..

I've been frequenting various forums for a long time, and this makes me realize how much added value it truly is.
What forums can produce, no video or social media can match. I was asking for help on which direction to take because I was lost, and you even wrote a guide on where to go. I thank you so much.

For the incredible quality of your post, but above all for the time you took to help me with my problem. I have no questions, everything has been answered, and now I just need to study and follow the path that I now know is the right one. My mind is now at peace

Really, thank you so much for your help and the equally amazing links. Best forum I've ever been on. I hope one day I can return the favor by helping someone here. Thank you all

GIZ

1 ·

Re: Hard Surface Blog: Sci-Fi Processing Station

Updated lighting with some rim lighting, some is real effecting all things in the scene, some is fake rim lighting. Utilizing channels in to affect certain objects and not others to help bring out certain details forward to the foreground with more light on it. This way I don't contribute to the indirect lighting as much or at all, and have a more complete control over the lights.

Solid recommendation, gotta say thanks again @Eric Chadwick it is definitely helping for sure. Trying to be careful not to remove the slightly 'moody' lighting look I was originally going for, but I think this is a nice balance for sure. Might dial some of the lights down in intensity, but overall pretty happy with the way it is shaping up. The little material sample cannisters, are now textured and in evnrionment as well. Which leaves one massive back wall asset to throw some detail too, and then work on different camera angles for beauty shots, and then a fly through.

Image: https://us.v-cdn.net/5021068/uploads/editor/pk/o1gab4yclzfc.jpg

mpuLse

2 ·

Re: Early-Late 2000s CGI/Game Models

Eric Chadwick said:

Darren Pattenden used to host a tutorial walking through his whole process of modeling and texturing a bust of Peter Cushing. I can't find the original anymore, but he does have this... "21 Tips on Capturing a Human Likeness" https://www.artstation.com/artwork/w8yrQO

I found the original tutorial on the Wayback Machine! I can see why he dropped it, he vastly improved the likeness in later attempts. But the workflow and insights are still golden! https://web.archive.org/web/20080714114321/http://www.daz-art.com/hi_poly_tut.htm

We also have some archives of the old now-defunct modeling forums Spiraloid and Subdivisionmodeling, see the PDFs here http://wiki.polycount.com/wiki/Subdivision_Surface_Modeling#Organic_Surfaces

Eric Chadwick

2 ·

Re: Early-Late 2000s CGI/Game Models

I remember watching Bay Raitt modeling gifs endlessly with my classmates back in the day.

Image: https://us.v-cdn.net/5021068/uploads/editor/us/607khqff2a8w.gif

Done in a software called Mirai
https://spiraloid.artstation.com/projects/KPyRB

jStins

3 ·

Re: Anatomy study entirely made on my Ipad

Well done on the model. Making it on an ipad is probably like making the Eifel tower out of match sticks.
If you are going to sculpt/model human anatomy its useful to have a great 3d reference on a screen next to the one you are working on. your second screen can be a really cheap one if you aint rich.

Daz3d is still free to use. You can change the reference model's height, weight, pose and also switch off the textures so you can get a better view of form. Rotating around a reference while you work is invaluable.

Image: https://us.v-cdn.net/5021068/uploads/editor/1i/bpomxh10gsyu.jpg

Also the pose you work in helps if it shows visual weight, a model is only a cloud of air 1 pixel thick. Pelvis slightly forward and the rest balanced though a cemtre line in profile. Legs slightly apart and feet at a relaxed angle will ad visual weight to your figure. Have fun!

kanga

1 ·

[PAID] Looking for 3D artist to create 3D art, based off existing concept.

Hired.

thexoplasma

1 ·

Re: Timeless Epoch (Procedurally Generated)

Sick!

Ashervisalis

1 ·

Re: Stuck: My Topology Journey Has Hit a Wall

@Gikkio Although there is some overlap in the basic art, observation, and problem solving skills required: concepting and subdivision modeling are different disciplines. While it may be possible to learn both at the same time it tends to be easier to focus on mastering each discipline separately.

It's generally recommended to practice hard surface subdivision modeling by recreating real world objects, with a lot of detailed reference material, instead of working from concept art that's vague or open to interpretation. This approach to learning from real world objects tends to make it easier for other artists to provide objective feedback that's focused entirely on the technical aspects of subdivision modeling.

Concepting can be done in 2D or 3D. Whether to start with one or the other could be it's own debate but one of the bigger challenges is that critiquing concepts and providing useful skill building feedback tends to take a lot of time and effort. So it's often more difficult to find artists willing to provide in-depth feedback on concept art. Especially when just starting out, because the supporting illustration or modeling skills also tend to require a lot of development.

When it comes to the existing concept: there's a few shapes and depth queues that could be developed further to visually suggest functionality. Both the concept art and subdivision model could benefit from further iteration based on comparison to or inspiration from real world references.

Something that can really help with building both of these distinct skill sets is to spend more time iterating at each stage of the project and analyze the results on your own, while also asking other artists for feedback, before moving on to the next step in the process. Take the time to really break things down and try different approaches or shape configurations and compare the results both to each other and similar real world objects.

At this stage of the block out the only real concern is testing which shape configurations have the most dynamic visual impact and plausible functionality. This basic block out could be iterated on several more times, by drawing additional inspiration from real world references and evaluating the model from different points of view (first person, third person, inventory screen, etc.)

Below is a quick example of a basic block out that follows the primary forms in the concept art and a second block out that further refines some of the shapes and adjusts the proportions around the grip and guard to match something from the real world.

Image: https://us.v-cdn.net/5021068/uploads/editor/yn/ej1kovdq1kvf.jpg

For the modeling process: consider spending more time in the block out phase and try to capture all of the primary shapes before adding any support loops. That way it's a lot easier to test different ideas without having to manage a lot of complex support loops and the final shapes can guide the loop routing.

Here's a link to a write-up about block outs and incremental progression for subdivision modeling: https://polycount.com/discussion/comment/2751240/#Comment_2751240

Below is an example of what an iterative block out process could look like. Start by defining key features like the barrel and slide then add additional shape details that create some visual separation between the surfaces. Create the basic outline of the frame and grip shape then use a series of bevel / chamfer operations to create softer shape transitions and round overs. Non-destructive workflow elements can be used to speed up the initial block out while also keeping things editable for easier shape refinement.

Image: https://us.v-cdn.net/5021068/uploads/editor/lf/0jchlsrsamay.jpg

Efficent subdivision modeling really comes down to keeping things as simple as possible while accurately representing all of the primary forms. The goal at this stage of the block out is to just create a basic mesh that has all of the larger forms and enough space around the shapes to create a clean path for support loops to flow around the shapes.

Here's a link to a write-up that covers how to manage some of the tradeoffs between mesh complexity and shape accuracy: https://polycount.com/discussion/comment/2751340/#Comment_2751340

Once the basic block out is complete, support loop paths can be added around the edges that define the larger shapes using basic modeling operations like inset or bevel / chamfer. Modifiers can also be used to streamline this part of the process by using edge weights and vertex groups to control the placement and width of support loops generated by bevel / chamfer operations.

After the primary features in the base mesh are locked in with support loops around the major shape transitions, move on to adding smaller secondary shapes and tertiary surface details, using the existing geometry to support the intersecting shapes whenever possible.

Here's a link to another write-up that covers how to analyze and plan out shape intersections: https://polycount.com/discussion/comment/2776197/#Comment_2776197

Keeping the base mesh relatively simple and using an iterative block out process to work from the larger primary forms to the smaller secondary forms and tertiary surface details helps keep the mesh clean and well organized. While also avoiding the time sink of having to re-work the larger shapes after the details have already been added.

Another thing to consider is that certain types of surface features can be added using "floaters" (detached floating geometry). Just be sure that there's enough room around any detached features for the flange that's used to blend the base of the detail into the underlying surface.

Image: https://us.v-cdn.net/5021068/uploads/editor/j1/msvgppms7ztr.jpg

Recap:

-Focus on developing specific skill sets independently of each other. Multiple skills may be required to complete a specific project but try to break the entire process down into digestible parts and seek feedback on each of the required skills before moving on to the next stage of a project.

-Dedicate more time to iterating on different ideas and shapes during the concept and block out phase. Try to accurately represent the shapes in each part of the object in a way that visually defines and clearly communicates their function.

-Block out the primary forms first, keep the mesh as simple as possible while still holding accurate shapes, then add secondary shapes and tertiary surface features in subsequent modeling passes. Let the larger shapes guide the loop flow around the model and use the existing geometry of the larger shapes to support shape intersections and smaller details whenever possible.

Some links to additional hard surface subdivision write-ups with animated workflow demonstrations:
https://polycount.com/discussion/comment/2779584/#Comment_2779584
https://polycount.com/discussion/comment/2777123/#Comment_2777123
https://polycount.com/discussion/comment/2780536/#Comment_2780536

FrankPolygon

11 ·