In this journey, we will also visit a newly built kopitiam (coffee shop) shophouse owned by a moderately wealthy family of Chinese immigrants, featuring subtle hints of Peranakan Chinese influence in its design and culture.
Your insights and critiques are deeply appreciated as I continue learning and improving this craft. I warmly invite all feedback from fellow artists and professionals to push this project to a higher standard.
╔═◈══════◈═╗
Links
╚═◈══════◈═╝
Personal Artstation: https://www.artstation.com/ongjianwei.
Project pre-production Document: Tanah Airku_Preproduction.
Project Tracking Board (Notion) : https://ongjianwei.notion.site/Headquarters-1f31557e1fd6805882b5d3bb6d36050b?source=copy_link.
Brief Development Logs (Notion) : https://www.notion.so/ongjianwei/Dev-Logs-2031557e1fd68073b0fddead6d882126?source=copy_link.
Formal Development Blogs (Tumblr) : https://jian-wei-24.tumblr.com/.
Replies
Rough CONCEPT ART
(The concept art are old with minimum research were done when drafting it out. Hence, the final design of the assets would most likely be different with justification and evidence.)
BRIEF EXTERIOR CONCEPT ART (Anushka Ghosh, 2025)
BRIEF INTERIOR CONCEPT ART (Anushka Ghosh, 2025 & Ong Jian Wei, 2025)
Further information about project can be found in Project pre-production Document: Tanah Airku_Preproduction.
Current Blockout in Unreal Engine
EXTERIOR BLOCKOUT (main focus will be the building in the middle. The buildings on the left and right will have minimum details)
INTERIOR BLOCKOUT
I focused on modelling and UV unwrapping the interior walls and floor, as the interior space is the first and highest priority area to be completed in the modelling process.
~ OPEN TO FEEDBACK AND SUGGESTION ^^ ~
Interior Wall & floor (Modelled and UV Unwrapped.)
Window Wall. Its upper Ventilation and main window are separated models from the window wall. (I am still unsure should I merge them all into one or separate them [If I were to separate them, I can swap out the window and ventilation designs for other building's modular pieces])
The window frame was UV unwrapped with attention to the natural wood grain direction.
Exterior appearance of the ground floor walls. The modelling of the doorway model and its supporting elements are temporarily halted to put more time on creating the interior furniture.
Kopitiam (Coffee shop) Table and Chair.
Temporary textures were applied to verify that the wood grain flows correctly, based on real-life reference images.
I'm currently developing the large cupboard furniture !
For the latest updates and in-depth insights into the modelling process of individual assets, feel free to explore the links in the main post of this thread.
I’d love to hear your feedback and thoughts !
The texel density seems to be a bit inconsistent (maybe exacerbated by the perspective).
The top of the table might be justified if you know for sure you'll stick with the softer marble texture or if it will be covered by a table cloth (doesn't seem like it in the concept, though) but maybe you just scaled the top as big as you could and then filled the rest of the space. The underside being smaller makes sense. The rim/side of the table is perhaps unwrapped as one piece, but it probably wouldn't be one continuous element in real life, either. So splitting that could give you some more UV resolution. You could even tile that if you haven't already, but then you have several tables and would have to be careful that the texture doesn't have elements that stand out too much.
For one table, you'll only ever really see about a third of the edge at any one time.
Edit: There seems to be a seam at the side of the table, so you perhaps split that already if that isn't the "main" seam (but then you wouldn't need padding for a continous piece that spans the full width of your texture). Maybe share your UV layout, if you want. After all, this could be a combination of different textures with different resolutions for all I know, and in the end, the UVs should be somewhat practical to work with as well.
Yea.. I am a bit conflicted on the texel density across all the current models as well. I will revisit this once I have a bit more furniture to see what shall be the common ground. (Do advice if this course of action is not proper or doomed to fail)
The table wouldn't be covered with a table cloth and will have a softer marble texture on the top~
I did split once on the side for extra UV resolution and also for realistic purposes of it not being one single piece. However, I only split the side once, would splitting it more than once be more realistic?
I will keep in mind about avoiding obvious wood texture repetition when I begin the texturing phrase!~
Ah, sorry! I have forgotten to attach photos of the UV layouts! Here are the UV layouts
Thank you again for your feedback, @Noren !
Deciding on / adjusting the texel density a bit later has the main disadvantage of you having to redo stuff, so I wouldn't push it back too far. So now is probably a good time since you already have a couple of elements. Obviously, the texture for the chair would be smaller than the texture for the table.
Splitting the rim more than once would mainly be for texture resolution. But the first thing I'd do is scale that part to span the whole width of the texture and then stack the other half on top of it and see if that's enough.
Likewise, I don't think the legs need unique UVs (unless you have decided on that for a reason). E.g. having 2 unique legs should be enough to avoid visible texture repetition within the object itself. If you have two legs next to each other and rotate that 180 degrees, we'll be looking at the other side of the legs. And if the texture is generic enough, you could even get away with one leg. Just make sure to move the duplicates to the side when baking your textures.
And like mentioned before, if the texture doesn't have too unique features that stand out and wouldn't repeat in the real object, it shouldn't be a problem between objects, either. If people really want to go and find texture repetitions, you can't stop them anyway. Technically, two limbs could have been cut from the same piece of wood as well and share the grain on one side.
Maybe you don't need texture space for the bottom of the table at all and can simply map that to the top if the material is the same?
If you need some pre-darkening with AO, you could do that via vertex colors (but then you'd have to find out if takes less or more ressources overall).
As for the layout, you could be more strict with putting identical/similar elements next to each other. As I understand it, you are trying to skip texturing the highpoly, so the orientation of the pieces matters. Also, you obviously already tried to group your elements, and sometimes moving them apart is the only convenient solution, but at least at a cursory glance, there seems to be enough space to alleviate this quite quickly in case of the table.
It looks like some of the edges don't contribute to the shape as well (bottom of the table looks different in the UVs, though, so that might be an older version).
You don't want terribly long and thin triangles, especially for bigger objects, but e.g. in case of the pillars on the legs, I'd probably remove the intermediate loops. Unless you need it for vertex colors or someone else says otherwise.
As for the marble being a texture with less detail and therefore requiring less texture space: Works for me, but you might still get questions if you show your UV layout, or worse, people might see it as a mistake (there's also the hypothetical possibility of someone asking you to change the texture later or someone else having to do it).
Roger that! Is this Texel Density alright?
Oh I had thought about cutting off the legs and have it duplicated for the other side but didn't come across of doing it. I have done the mentioned changes. Using the image above, The back two legs are a separate model using the same UV.
Ah the top and bottom part of the table would be different material. The top would be marble texture and the bottom would be wood texture.
I am sorry but I am not knowledgeable about pre-darkening with AO.. Do you have any external websites I can read up about this? (If you don't, no worries! I will look it up later)
I have deleted intermediate loops~ There isn't any plans for vertex colours for the tables as the furniture would be new because its a newly built shop. Hence, there wouldn't be too much obvious wear and tear or damages to the furniture. The remaining edges I kind of need them to keep the roundness of the shape.
The current marble texture seen in the previous reply is a place holder as I have not reach that texturing stage yet as there are many other assets that are yet to be modelled x.x
Current updated UV Layout!
(I make room for the chair UV Layout so i can duplicate another chair and has its UV take that empty right slot. Then after texturing the two chair as a single object, then i will separate the two into two unique chairs using the same material [I hope I am making senses >o<])
Sorry if I’ve misunderstood and done something different from what you were referring to. I really appreciate your patience and support! Thank you again!
Also, for now, this only tells us the relative density, not the absolute one. To check the actual density, you'll have to work with pixels. Probably can be inferred/calculated from your checker pattern for various resolutions, but it's better to go with a pixel sized pattern directly.
Assuming it's not an illusion due to the perspective, the rim still needs a higher resolution in my opinion. There could be a rationale, like e.g. the rim doesn't need normals and the feet do, but I'd try to get this a bit more homogenous. And if you haven't nailed down your final texture for the table top, then I would give that a higher resolution, too.
I can't tell 100% what you are doing right now with the rim, but like said, I'd use the full width of the texture and/or maybe even tile this some more. You could also detach the backside and scale that down, but you'd have to decide that while actual having it in front of you (is the gain in texture space worth the inconvenience, maybe you'll split it anyway for normals etc.).
And unless this is for a shooter, where people crouch around or topple over furniture, the underside of the table will hardly be seen, if ever, and even with a different material, the underside "steals" too much texture space from the things that will be seen the most and closest (like the top and the rim). You could half or quarter the UVs, here (pie slice like), especially since there are already beams underneath that would hide the seams conveniently. Or you could do a generic wood texture and squash your UVs (very situation-dependent and not artist-friendly, though).
You could also model the underside as a separate element with fewer segments that just sticks in there. You won't save on vertices since they are already there because of the rim, but this would be a case where there's not need for a lot of long thin triangles. Would it actually matter? No idea.
I don't think that's a fixed term. I was just referring to baking an Ambient Occlusion pass (or some shading in general) into the diffuse/albedo in case the visual quality of the render engine or settings isn't sufficient for your intended use and you aren't using light maps (that often have a second unique UV set, calculated automatically).
What I would probably do is set the texel density, then, ignoring the 0-1 UV space for this step, scale the elements in such a way that most have the same relative density minus the stuff that is hardly seen or farther away from the camera (which technically includes the legs, but maybe you want to show off your handiwork in some closeups). Then decide on a texture resolution for this object and arrange the UVs accordingly. There might be some overlap or back and forth between these steps as well. If you have some free space in your UVs, and the large islands are maxed out, you can scale up smaller elements, but usually there's nothing gained by that other than giving the illusion of a tighter packing and there's always the possibility that you'll need that space in the future.
I hope this isn't too confusing. I know I tend to ramble and offer a lot of caveats.
The planned environment is going to be a first person viewing and allows audience/players to roam around. I don't plan any crouching movement or interaction at this stage of production ( Most probably will implement it when I am done with this project submission by polishing this into a portfolio piece)
The checkers map I used is generated by blender by making an UV grid image. I did the settings of 2048 x 2048 px. This a 1 m dimension cube with the UV grid texture applied along with its details with the Texel Density Checker addon in Blender. I plan to keep most of the objects to be 2k textures and 4k for more complicate objects or objects that will be viewed closed up.
This is the new UV Layout and an orthographic view of the table. The rim and surface top are 9.784 px/cm. Any higher the rim and the surface top would exit the grid.
UV comparison with Chair and Table in orthographic view.
I apologise if you had to repeat any steps or techniques in your feedback that I still have not visibly implemented or reply to. Please feel free to point it out, as I may have unintentionally overlooked it while rushing or got confused amidst researching the terms too fast until I had forgotten. > ~ < !
Thank you again for your feedback and patience!
9.784 px/cm sounds reasonable as well, but personally, I would still slap a checker map with 1024 tiling on there (so 2048 pixels) and move around a bit to see what that translates to in praxis (not really practical to post screenshots of that, though, that's more for yourself).
The edge padding might be a bit small in places, but that might be a wrong impression due to the screenshot resolution.
Here's a link just in case: http://wiki.polycount.com/wiki/Edge_padding
And you can use less if none of the maps has any real contrast in that area (more tricky with normals, though, at least with different island orientations).
It's Noren, btw., not Neron (no worries, though).
Alrighty! A checker map? Would it be the one that I have been using or the colour checked map? Also the checker image is generated to be 2048 by 2048 pixel.
Oh! i have not heard about this before. Thank you for your wisdom! I will do some editing to the UV to leave a bit more room for it.
Thank you for your feedback, Noren! I am currently working on the cupboard model. Should be able to post an update on that by tonight!~