Medieval Stronghold Architecture

UPDATE 12-07-2018:

I think I’m almost ready to push this to Epic’s. The final render is done! On top of that, the current scene run at 120FPS everywhere on my side in-engine 4k with a 1080TI, with a polycount going from 500k to 2.5m and draw calls from 500 to 1.5k in-context. It seems like the optimization pass is helping a lot more than anticipated!

Some stats: When englobing the scene (which means the camera is far away and only look from a somewhat top-view, you would never have this PoV in a game without fakes), I’m sitting at ~5m poly and roughly 90FPS.

The double-sided set has 11 modules. 6 has been crafter for exterior meshes, 5 for interior. They’re very similars and should only be used if you want to let players build their own bases, or when you have to implement large, uninstanciated modules in a dynamic lighting setup.

The Scaffold meshes and modules were done back from scratch. The base planks used for these used to have a lot of wasted polygons. I’ve redone then from the ground up and gave a good optimization pass on the ropes. Small warning: To keep consistency throughout the whole Scaffolding setup, I’ve had to change some pivot points. This means that when updating, you’re likely to see modules that have a rather large offset. I’ll have the full list when the patch goes through!

I’ve gone and revamped the lighting of the scene too, to give it a better feeling and better control on screen brightness when navigating.

Of course, there’s a lot of bug fixes mixed in this. Floating modules in the demonstration scene, collisions that have a lot of wasted polygons, etc…

Finally, I’ve remade the interior walls’ textures in 4k, as per requested. All the interior modules have been updated to reflect the new texture – you should notice a lot less of artifacts!


UPDATE 11-29-2018:

[spoiler]So I’ve been hard at work these past few weeks to give a little overhaul to some elements of the pack, in addition to crafting some two-sided models. Here’s a small preview:

As you can see, it’s almost complete. I have some minor bugs to fix first though and perhaps add some more two-sided modules. The lighting has been entirely overhauled in the demonstration map so it’s easier to see in the interior areas - and more beautiful too. I’ve remastered the interior wall modules’ texture from 2048 to 4096 as it was requested a lot. Plus, I’ve edited the normals so it’d look more like a rock and less glitchy in some areas. Plus a lot of little tweaks on materials here and there.

I’ll probably have some baking to do when I do the last few bug fixes, but it’ll be sent to Epic before the end of next week![/spoiler]


It’s finally live! You can access it from the official page here:…d-architecture

Here’s a quick tour of the demonstration map in addition:


[spoiler]New this week (July 26th 2017):

This is the last week I’m spending full time on this set. I have received some requests to add some modular pieces and I’ll try to add as many as I can. Until everything is done, I will not do another capture of the demonstration level. Instead, this week, I’ll spend more time on the technical aspects of the pack - some of which are of interest to level designer, level artists, props artists and - to some extent - programmers.

First things first, I’ve made three videos this week. One is about showcasing the showroom, where all the modules and props are exposed individually. The other two show how you’d build your modular architecture with a real-time building, for the interior and the exterior. They both are not taking every single piece of architecture, there’s just too many of them for a small video of a few minutes each. Still, it should give you a clear idea of what the pack can and can’t do.

Now, on to what I believe will be the most asked questions.

Q: Any ETA on a potential release for the pack?
A: Not quite. I’m still spending this week polishing and adding last-minute modules, then it has to go through Epic’s validation pipeline. It should then be added to the next Marketplace release, which I suppose would be by the end of August.

Q: Have you thought of a price?
A: Unless something major happens, I plan to sell the pack at US$74,99.

Q: Is it optimized? It seems a lot of detail went into the modeling so I’m a bit worried about performances.
A: You’re right that the modeling has a lot of triangles. But that’s not an issue when there are several LoDs taking care of optimizing everything. For instance, if I take the whole scene in my viewport, I get about 65~70 constant FPS (You may be interested in taking a look at this picture: ). One important thing to consider with these numbers is that since it’s taken in the editor - IE not in play mode. So vital optimization mechanics like mesh culling are inactive. The performances outside the engine, like in a genuine game build, would be marginally better.

Q: Great, so there are neat LoDs. But that didn’t quite answer the question - what can be expected at run time?
A: Better performances. For the time being, the scene run at about 65~70 FPS when in full screen (1440p) during ALT+P with a GeForce GTX 980. Although again, it’s in-engine and so you have a lot of background processes going on at the same time. This ( ) screenshot was taken at the most “inefficient” area of the map where mesh culling is somewhat inactive. As soon as you move indoor, the FPS boost to 70FPS+. AA, Bloom and Motion Blur are also activated.

Q: What about the textures?
A: Right now, the scene uses baked lighting. Unreal automatically cap the texture memory to 1000MB, so as soon as this is hit you get a message warning that you’re exceeding the limit. This is not the case with the scene. On the 1 000MB budget, 780MB are used. Again, this includes baked lighting, which generates several thousands of smaller texture maps. There are ~ 3500 actors in the scene. Remove maybe 50 and you have a number of lightmaps being generated. And yet, it’s all fitting within the defined budget and then some, so you should have no issue implementing your characters/props/UI/etc.

Q: I’m sure I saw some vegetation in your screenshots. Why are they not on your stats screenshots?
A: Because they’re very heavy. This pack is not intended to be bought with the intent to have working vegetation. It’s an architecture pack. I’m still weighing if I keep them in or if I remove it altogether.

Q: So what would you say this pack is aimed at - what platform(s)? Even if it’s well optimized, the stats exposed are not what is considered “light”.
A: For Windows/PC gaming. With some light tweaking work, PS4/Xbox One. Definitely not VR or Mobile, or at the very least not without a sizeable experience with the limitations related to these devices. Anything running on modern hardware, basically. Then, with proper settings (removing AA/MB/Post-Process and making LoDs + MIP swap faster) the pack can run without issues on less performing PC hardware.

Q: How long will this pack be supported?
A: For as long as it needs to be. Like the other pack I’ve worked on - Dwarf Fantasy Environment Set, which received an update this summer with a complete overhaul of its textures - I plan to update the pack according to new technologies and upgrades that can substantially make the pack better.

Q: Will there be something included to help me get started?
A: Absolutely. I’ll join a thorough PDF guide on how to use the set, potential pitfall and what to pay attention when using it.

Q: Can I create custom assets? I know how 3D modeling work a bit and I’d like to make my own stuff with your materials.
A: I’ll join the “base” meshes I’ve used when I was crafting the set. This should allow you to make virtually any piece you want. The only limit is your imagination (Not really, texel ratio and polygon count are but hey, don’t let me shatter your dreams!)

Q: How many assets does this pack have?
A: I don’t have the precise final number as I’m still adding some stuff. But as of right now, there are 174 meshes (excluding the 28 terrain assets specific to the demonstration map). From there 174 meshes, 9 are furniture, three are vegetation, three are flags with PhysX, one is a canvas for scaffolding if you want to make your own meshes. For the remaining 159, 12 are optimized merges of smaller architectural meshes to reduce a number of draw calls.

In term of materials, there are 22 unique materials and 13 material instances.

For the textures, there are 110 unique maps. I still need to go through and remove those that I no longer use, so 100 would perhaps be more accurate.

Q: Anything else I should know?
A: The pack was made with a baked lighting pipeline. While it’s certainly usable in a dynamic lighting environment, you’ll need to do additional steps, like making map-specific fakes and dynamic loading that I simply cannot do without direct access to your specific project. That’s really important to understand, just like everything using dynamic lighting, that you’ll have to take additional optimization steps.

New this week (July 19th 2017):

So many things to show. In fact, there’s so many that I had to retake all the screenshots altogether for the exterior (and some parts of the interior architecture) - You can access the high resolution version (1440p) of these screenshots here:

<–Snipped for readability–>

The architecture itself is 95% finished. There’s a missing module here and there to add, some lightmap fixes to do, but what you see here is what you get. I don’t have any plans to change them any further, but if I have enough time, I may add additionnal requested modules. Next week, I’ll have a complete video of the demonstration map and the overview map, so you’ll be able to see in real time how it looks, feel and react to the camera – plus, you’ll get a thorough look at how the modularity works.

I’ve started working on the furniture set. It won’t be anything too fancy, it’s mostly to give some depth to the interior stuff. There’ll be better lighting with torches by next week. I’ve also worked on a shader that uses a baked AO on unique furniture pieces to “hide” the fact that it uses a modular set of furniture pieces. This is how it looks like in the editor:

I wanted to show it in the scene with baked lighting, but there are some issues with the modeling that gives it unrealistic lighting. So I’m working on that today.

For those who wonder about the target price for the pack, nothing is set in stone just yet, but I’m undecided between $69,99 and $79,99. Excluding the terrain, there are 143 modules in the pack plus a few more missing. There’ll be at least ten furniture pieces, 93 texture maps and 20 materials. I’ll also throw in the basic assets used for the modules, so you can craft your own unique pieces if you need to and have some knowledge with 3D softwares. Let me know what you’re thinking!

PS: I think this deserves to be said and shared: The shader used for the grass is almost directly taken from…y-Grass-Shader - The modeling is made by me, but not the shader itself. So in full transparency, you can access that link to check out the post and access the scene with the free shader!

EDIT 7/11/17 - Showroom video showcasing current modules:

New this week (July 10th 2017):

Started working on the interior architecture. There’s a lot of things that remains to be done (fix some bugs, tiling issues, baking issues, will rework the doors again), but the backbone is done.

I thought that maybe you’d like to see how the set works in real time (mostly interior), so I recorded a small two minutes in Fraps.

What I’d like to do this week is to basically finish the interior architecture (towers still needs to be done) and start working on the terrain outside. Then next week, it’d be mostly polish and, if time allows it, a small modular furniture set for the interior. Right now, there’s well over a hundred architecture modules and I expect this to keep going up, perhaps close to 150 by the end of development.

Slight Edit: Added corrected LoDs

Good day everyone,

I’ve been working on a medieval architecture set these past few months and I thought it was finally time to expose this on the forum. Those who follow me on Twitter/Facebook probably saw some pictures of it here and there, so here’s a more complete rundown of the current state of things – but first, some screenshots (high resolution album available here ):

** Access the high resolution album. With new updates, removing the screenshot clutter is necessary! **

So as of right now, the complete scene is populated by 2.5k objects, originating from 70 modular assets. The light is completely baked on Production quality with no signs of trouble with texture memory. There is a total of 14 materials and 47 textures.

With a 980GTX, I run the scene in 1440p at 70 FPS when having the whole thing in camera view. When moving around, the FPS move between 60 and 90. As of right now, there has been no mesh merging, which mean that the amount of draw call is high – but since it’s a demonstration map, I probably won’t do it. In a context where the pack was used in a game, it would be highly recommended to merge similar assets together that are neighboring each others and then do a LoD generation. The meshes are also “omnipresent” for the time being, which means that they don’t “disappear” after a given distance. It’s something else that would greatly benefit the optimization of the pack in a game context, but which is useless in a demonstration map.

Each modular piece have several LoDs, which allows thorough level of precision and quality at any point. Example:

** Pictures has been removed for better readability. Full album with all LoDs at and **

This allows the pack to bolster modules with several thousands of polygons without negatively impacting performances. In fact, the scene has roughly 2 million active polygons when entirely seen by the camera, whereas if forced to their basic LoD, the scene has a whopping 25 million polygons. During active gameplay, the count of polygons is moving between 800k and 1.6m.

With that said, here’s what remains to be done with the pack:

• Create special decals for ashlar walls, to break the visible patterns
• Add 45 degrees (vertical) modular ashlar assets
• Create the whole modular interior architecture
• Add wooden support for structures
• Add courtyard terrain

I’ll keep you updated as much as possible in the following weeks. I’ll try to answer your questions ASAP too.

PS: If you’d be so kind, you can also follow on Twitter and Facebook - and - but obviously it’s not a request. If you find the work here interesting, a little support can go a long way![/spoiler]

It looks fantastic, how much time invested so far?

Little curuous about the lowest LOD for the stairs, seems a bit aggressive with the reduction does it not? Changes the silhouette and tears faces apart, seems harsh compared to the LODS for the door? Looks Almost as if it was meabt to be a mesh designed for multiple levels of destruction.

No offense everything else included in the post itself seems to look great

I haven’t kept a precise count on the amount of time invested, but saying there’s roughly a month and a half in it would be accurate. I plan to spend about another month in it, so it should release by the end of this month or next.

Yes, you are right about the aggressiveness of the stairs vs the door LoD. The main difference is that the stairs gets rendered more often at larger distance, because it’s an outdoor module, whereas the door will always be integrated inside a wall – the engine will not render modules that are not seen by the camera. Likewise, LoD4 is really far from the main camera, so there’s no actual popping (even if the mesh is utterly destroyed).

With that said, I think you’re right and it’s too aggressive. LoDs are not final, so I’ll make sure they at least keep their silhouette.

Wow, this is really nice. I am going to buy it for sure! :smiley:
I already loved your dwarf environment set, really good work with this one.
Allow me to ask: Will we see more medieval assets from you? Or do you plan to make this asset the last one? If possible, I would like to buy more stuff with your art style.

PS: It may just be me, but the texture for the door looks a little too shiny or low-res compared to the rest.

Well it all looks great thats for sure i was merely curious, glad it isnt mostly just my technical ignorance or something lol.

I hope someone puts this to good use, medieval and egyptian / sandy desert environments and architecture were some of my favorite styles to work with back in UT2004s golden years

Haha I’m glad you like it! And your support is really appreciated.

I will do other sets, that’s definitive. It won’t happen before next summer (Summer 2018), because University (Computer Science) eats too much of my time. Still, I’d like to work on an Elvish set next, to complete the “Human/Dwarves/Elves” trinity.

About the door, yes it’s way too shiny. I’m currently working on the doors’ textures as there are a few things I’d like to fix, including the shine and the “pixelation” on the sides of the wood planks. It’ll be there on the next update.

The polycount seems quite high but your textures don’t look high detailed as well when you move close to the objects. Are you using PBR, LERP, Vertex Painting ? 2048x2048 or 4096x4096 textures? I mean you door LOD0 is 75k+ polygons so it is an high poly model for an object like a normal sized door and still I see a lot of detail (high frequencies) on the metal ornaments only while the wood textures look washed out lacking detail. The walls bricks textures I don’t know… you didn’t add a close up picture but if they are like the wood ones then you need to improve those too. At a distance everything looks excellent but an high poly modular set like this one needs to look very detailed everywhere and I don’t think you are going to set a cheap price, right ? The 2Million+ polys example castle you show can’t have the textures washed out anywhere, it would just be unacceptable if you are going for an high price… it must look excellent for FPS gaming as well and so moving very close to objects.

Hype: “Create a modular interior architecture”.

This is gorgeous, great work!

Hi Dark,

Let’s delve in the technical first. The textures are mostly 2048x2048, based around specific meshes (stones, wood planks, ground tiles) that can then be used at will in modular pieces. So, for instance, the walls uses two textures: One for the bricks, and one for the plaster layer. The bricks uses a 2048x2048 map while the plaster use a 1024x1024 texture. This allow larger meshes to have a highly defined visual without costing too much in term of GPU memory. The only exception to this is the brown wood (not the doors), which uses 4096x4096 as it is made of 18 unique 2 meter pieces.

The materials are all PBR. They’re mostly crafted in Quixel DDO, which uses scanned materials. Obviously, they’re highly tweaked on my end as using brute materials wouldn’t feel very good visually.

To answer your question about the wood material looking pixelated, this is a result on my end for having tried something for the borders which turned out to not quite be what I had in mind. Here is a small album on imgur of some close, high-res screenshot taken during gameplay:

As for the price, it has not yet been decided. What’s sure is that it will not be below USD$50, but what are your thoughts about a potential price?

I just want to clear something else, which I should’ve said in my first post: This is made as a third person asset pack first and foremost. While it certainly can be used for FPS games, it is not being crafted with this in mind. Even less so for VR.

Hope I was able to be able to answer your questions!

I dread that moment. It’s incredibly intimidating, but I think it’ll be worth it.

Coming from you, that means a lot. Your work is astounding in itself, so thank you!

Updated with this weeks’ stuff! Look at the edited first post to see what’s new.

Very interested in this!

Only thing i suggest to change, the repetitive wall design of these tower walls.

It looks great, my only feedback would be to add some color into it. Everything blends in. Just some examples but the roofs blend into the tower bricks, the bars in the windows are using the same brick color. In the video, even the wood doors/parts blend into the brick parts.

With that said, it looks awesome and I’ll pick it up. It just blends together.

That indoor shot is fantastic. Great work on this pack. Get some cloth flags blowing in the air, it will bring life to the environment. :slight_smile:

Also at that spot maybe some kind of assembly hall with bluish windows or some borders at least, and a larger room? Its high up on the castle, thus a super room … and a balcony would be great here, with wood pieces even. The perfect spot for a cinematic scene, from where the king talks to his soldiers in the court …

These outstanding wall pieces should be aligned to the front.

The castle entrance could come with a bridge entrance similar to this Epic_of_Tales_Castle.jpg
Also notice these wall structures at the entrance (first image)

That’s something I’ve been pondering this week. The problem with these towers is how they work with regular modules – because the goal is to prevent tilling as much as possible, the modules are not fitting into a precise 2x2m plane. Well, the plane behind the stone is so you never actually have clipping issues, but the stones themselves leak on both sides to break the repetitiveness. This leads to some issues with angular modules, like this one, where a module has to be patched on these corners to hide the “floating”.

Now what I’m thinking to do is to create a specific 2x2 module just for the towers (one with a 45-degree angle, the other regular). This would remove the need to use these patching modules, although I’ll still leave them in the pack for those who like them.

For the time being, the set will probably keep the same shade to blend everything together. The reason behind this is that most pieces of architecture are exposed to the same climates. Sand, dirt, and moss all take possession of the materials used in the stronghold over time – The goal would be to break this by using outstanding props, NPCs and lighting to add some kind of dynamic to the scene.

You can be certain there will be flags. This time I’ll try to make them work with Physx, to give a really neat look.

Got to say that you really know how to envision a narrative. I do admit that this sounds really great.

Now, the thing is that this part is what we would call as a fake. This is something that was done throughout the middle ages, where castles and strongholds are purposefully inflated beyond utility. To give you a practical example, the Hedingham castle in England had, from the outside, five layers. A basement, a lower-level ground and an upper gallery. The two top layers were empty and made of fake holes to give an impression of grandeur!

I am unsure why, but I am unable to open your attachment.

I’ve been thinking about adding a bridge. That’s something I’d definitely like to do, but I don’t know if I’ll be able to fit it in there. I’ll definitely do my best to do it, though. Thanks for the thorough feedback @unit23

And thanks for your feedback to my input, i am generally a very critical person (consider me a painter). Cheers

Direct link