My Initial Thoughts after 1 month using UE 4.24 for ArchViz

I run a small architecture studio and we produce everything in house, from project to presentation.
My desired goal would be to migrate to a Real Time RayTrace high quality renderer/engine and improve my workflow and output speed, and also improve de design iteration process.

My current workflow is going from rhino to Max and then render in Corona. I like Corona, it is simple, fast and produce good quality images and has the advantage to be integrated with 3dsMax which has a lot of great plugins like Itoo forest Pack.

I’ve been experimenting and learning Unreal engine for a month and I would like to share a bit of what I’ve learned so far, what I think are the pros, cons and things not working.

But first, my test setup:

  • Nvidia 2080ti RTX running studio driver
  • Threadripper 1950x
  • 64 GB Memory
  • Unreal 4.24

First lets start with initial WORKFLOW:
Import and initial Setup:

RHINO TO UNREAL
The best way I could figure out to import my models from Rhino to Unreal is using the DataPrep plugin, it’s promising ;). It allows to automatically assign unreal materials to the assets, swap placeholders assets with already setup actors inside the unreal project and other functions. Great time saver.
[INDENT=2]
PROBLEM - The datasmith is not importing the UVs from Rhino, so I need to convert everything to meshes inside Rhino and then import. :([/INDENT]

The Dataprep generates an asset called “Dataprep_Rhino_Scene” which allows to update the scene if any changes to the original file is made. It will reimport everything though and erase changes.

3DSMAX to UNREAL
I used the datasmith max plugin to export the scene assets to unreal format. It takes some time but the assets comes out as it should.
[INDENT=2]
PROBLEM- Materials and textures conversion from Corona and Vray to Unreal is not working properly, if you have a color correct node and real world map size in max it will bring some weird textures and also a lot of functions are not supported or converting properly. A lot of rework here, specially for vegetation which comes without two sided foliage translucency. :frowning:

[/INDENT]
Observations:

  • I don’t know yet a good workflow to import/re import a single asset like a chair or a scene asset set. Haven’t tested yet.
  • As far as I understood, Unreal creates a new file for every asset inside the unreal project folder. So, every single texture will be converted to a new asset. There is a drawback here since it will probably consume a lot of disk space.

SCENE MANAGEMENT:
MATERIALS:
[INDENT=2]I created a master material trying to mimic the functions I have inside 3ds Max/Corona and them having a library of instances for my most used materials.
The end material is kind of complex with color correct parameters, normal map or generate normals from a texture, detail normal, use real world UV map size, rotate and flip textures, parallax shader, displace, override parameters for raytrace optimization.

Features to be added to my master material:[/INDENT]

  1. Round edges parameter - IT IS A MUST for Archviz but it seems to be very complex to do, I HAVE NO IDEA HOW TO DO IT
  2. Randomize the UV by asset instance and elements - I HAVE NO IDEA HOW TO DO IT
  3. Triplanar mapping
  4. Dust from world coordinates (or something like that)
  5. Dirt map to generate variation
  6. Edge weathering
  7. Wetness
  8. Dirt by proximity to other assets

[INDENT=2]
PROS[/INDENT]

  1. Quixel integration is amazing, they have some really good materials and textures
  2. Realtime blend material by vertex painting is SO POWERFUL, haven’t tested a lot yet but the videos from Quixel team are just amazing
  3. Material layers seems promising also, I need to test them
  4. Decals are also great, it allows to add variation without the need to create complex custom materials
  5. Parallax is great for adding details without using displace

[INDENT=2]CONS[/INDENT]

  1. The blueprints are powerful but I kind of missed an all purpose material like in Twinmotion or Lumion
  2. I also missed ready nodes for color correction, random Uvs, round edges, adjust intensity of normals and displacement. I had to figure this out, which is kind of time consuming, specially for inexperienced users.

ASSETS MANAGEMENT:
[INDENT=2]I need to do more tests in this area, some pitfalls I could not figure out how to solve is:[/INDENT]

  1. Adjust mesh pivot point. I know how to do it per actor placed in the scene but I don’t know how to do it in the main mesh. Don’t know if it is possible
  2. Add and manage new material elements on meshes. I could figure out how to do add a new material element via the Mesh Editing plugin but then I don’t know how to delete or change the material elements.
  3. Paint foliage tool is great to scatter assets but it lacks a way to manage the added assets in the “scatter window”. It can become kind of complex really fast.
  4. As said earlier, Unreal creates copies of textures inside the project folder instead of reference an outside asset, so a 1.5 GB 3ds Max file bacame a 8.5 GB project in Unreal after using the datasmith plugin.
  5. Don’t what is the best workflow to easily manage and add assets from my libraries of furniture, plant, etc. Mostly .max files.

RENDERING: [INDENT=2]
As said earlier I’m trying to make everything ray traced, no light bake. Here are my impressions so far.

Overall the ray trace technology in Unreal seems promising, I now it’s everything new and under development but my guess is that it will evolve to be consistent, fast and with great quality in the near future.[/INDENT]

  1. The ability to setup exposure, depth of field, lights in real time is great and a real time saver.
  2. The cinematic controls are working very well, easy to use, easy to learn, consistent. I could not output an alpha channel pass using the Sun Sky system tough. And I saw some complains about stability to output large formats.
  3. PBR material workflow seems to be working nice
  4. Lights and shadows are consistent

[INDENT=2]**RayTrace Reflections - **Are working fine, roughness maps are consistent. Depending on the setup it will be noisy or loose performance but that is probably limitations of hardware and early implementation.

**RayTrace Global illumination - **The new Final Gather is great to work in the scene and also output images, fast and if you wait some time it can get rid of most of the noise. The brute force seems to require a lot of processing and becomes unpractical in complex scenes.

RayTrace Refractions - Seem to be broken, :frowning: I spend a lot of time trying to figure out how to do a simple planar glass. I did it setting up the material specular value to “0” and setting the “Max. Refraction” in the PostProcessVolume to “1” or “2” anything more than that seems to break everything, For exemple, if you are looking thru a glass panel with the Max Refraction setup to more than 2 it seems to “cull” some features like Ray tracing shadows from the Skylight and also it makes the lower LOD grass disappear. Really funky
Also I could not figure out how to do the border of the glass to look darker as it should.

PROBLEMS:[/INDENT]

  1. My scenes do a really annoying flicker after turning on the translucency. I could not figure out what is happening. the only way to solve is to close thw engine and open it again. But after some time it will show the same problem again
  2. I’m trying to use the Harry Emelianov Volumetric clouds but the element turns black when using translucency
  3. I Could not figure out how to use Two_sided_Folliage with Raytrace, it works fine with raster but when Raytrace is turned on it disapears os seems to be really, really small on the borders.

[INDENT=2]Don’t know yet how the render will behave in interior scenes illuminated only by lots of artificial lights.[/INDENT]

Final considerations:

  • Overall Unreal seems promising and very powerful, it’s also complex to learn and implement simple features that we are used to in archviz, a lot of digging is needed.
  • I believe that if the Unreal team is willing to be a consistent tool for archviz, a lot of presets and simplifications are needed. Ray trace is not ready but can already output some good results depending on the scene.
  • Being able to directly import a .max file would also be good since it will allow us to import assets from libraries without opening 3ds Max.
  • I would really like to understand better the plans for Twinmotion and Unreal, since I don’t know which one to focus my learning. Right now the actual version of Twinmotion does not create a quality good enough but it has some cool pre built tools and ready to use assets.

That’s all. I hope this can help someone and also create some good productive conversation about workflow and features.:slight_smile:

3 Likes

Thank you for posting this, many peoples will find it useful. :slight_smile:

Nice post. I too have been trying to work with Unreal in archviz with some test projects. I have been mostly using Enscape which is just too easy to get pretty good results rather quickly in real time versus Unreal. However, for some reason, I keep trying to learn Unreal and seeing where I can go with this. I did the same ■■■■ thing teaching myself 3dsmax after having really mastered sketchup. 3dsmax is so much more complicated to accomplish what one needs for archviz work but yet I keep on using it and teaching myself new things. I guess I just enjoy self torture that way… If it was not for Datasmith, I am pretty sure I would have given up on Unreal because of all of the time it would take to unwrap and UV every stinking object in a scene. Datasmith has done pretty well with this but it is still not optimal and technically, if you want the best results, you still should be setting up the assets UV’s in 3ds max to get them right.

I can’t respond to all of your pro/ con comments at the moment but they were good. However, it seems like you are saying you do not have a process for bringing in individual assets into unreal from 3dsmax. I am not sure why you just do not use Datasmith to do that. If doing a chair, I set up the chair model the way I want with standard materials (vray usually works now too if using bitmaps and normal maps), make sure my UV’s are set up, center the pivot on the object, select the whole chair and then export via datasmith. Once in Unreal, I import through datasmith and then I can open that asset and edit its materials by either editing the imported material nodes or creating a new material in unreal and dragging it into the static mesh material slot.

Seems to work okay for the most part but it can get involved with how it needs to be set up in 3dsmax. My problem now with 4.24.1 is you cannot do a package build of your project to create an executable. It seems others have had this issue as well.

Thanks for the answer,

I believe you are right, to import assets from max the best way is really datasmith which creates a folder structure with the assets. What I don’t know though is what is going to happen if, for example, I export a chair with the name “furniture.datashmith”, now I want to export a table using the same name “furniture.datasmith”. Will I have to erase the previous folder? If so, Will my unreal project lose the reference to the previous asset?
I know it is a simple test to do, I just haven’t tested it yet, so, I’m not sure the best process. One thing is for sure, I don’t want to export every asset with a different datasmith name to save disk space.

The other point about this is that it would make more sense to import .max file directly from unreal, or have an external converter of some sort. This way there will be no necessity to initiate max , open and export each asset. It’s a time consuming process that does not add value.

Thanks for sharing that. I am going to put out a thought although I know you guys won’t like it. The main problem is that you want to use your old workflow on a new or different technology. Let that old 3ds max / corona workflow go. Instead of building the scene outside unreal you should build it in unreal using unreal assets. You may not have many unreal assets right now but that is where you should start. Build a library of unreal assets made the way they have been made for countless high quality games. It feels like you are putting bandage over bandage onto your old assets to make them work but they really never will. I’m saying that from my own experience having worked in an ArchViz-studio for a couple of years. The boss always thinks his assets bought for $2 from russia will work fine but they don’t. Every hour you spend trying to fix them is a waste of time you could spend building a better asset from scratch. - Sorry for being negative but I feel I just have to put that out.

I believe in you, my idea is to have an asset library inside unreal but that takes time and I have a huge library, so it is time consuming and a long term process. In the meanwhile a blend process is needed.

Actually, I am going to disagree with this for now but only because Unreal is not set up to be a modelling program. There is now way it makes sense to build up and archviz scene in Unreal. For the AAA games you mention, I am pretty sure I am correct in saying that many of them are creating their assets in an outside modelling program whether it be Maya, Blender, or 3Ds Max. There is no other way to do it. There is a reason Unreal just threw a boatload of money at Blender and there is a reason there has been alot of time and resources put into Unreal Studio (Datasmith). Unreal alone is not an effective tool for archviz project from initial concept to final product. No software that I am aware of currently is. I use alot of them including Revit, 3dsmax, Sketchup, Vray, ZBrush, Enscape, Photoshop, and now Unreal and some Unity. Each one of these is a tool at this point used to accomplish a certain task for a specific purpose.

With Unreal, the focus is on realtime rendering and then being able to output results that can be used in interactive environments on multiple platforms. Creating the 3D assets is still best done with other programs and then using the FBX export/ import method with wrapping/ unwrapping UV’s or using the Datasmith approach which is trying to streamline this tedious process.

You are right. Unreal isn’t a modelling software and it never will be. What I meant with building the scene in Unreal is to take all the individual pieces created in Blender or Max and put them in place in Unreal. For me an ArchViz scene has multiple types of elements.

You got the building itself which is always unique and can be brought in from CAD or Blender. Most of the time I would model the model it from scratch. The geometry is simple and the result is clean while any converted mesh from CAD can be very messy and it may take longer to fix the geometry than to rebuild it.

Then you have the furniture and decoration which is reusable and doesn’t need to be recreated for each scene. These assets can be in your Unreal library. If you have a couch that is already in your library it takes less than 1 minute to place it. If you have all assets you need dressing a living room can take only a few minutes.

Then you have some sort of in-between elements like doors and windows or the kitchen cabinets. Doors can be reused most of the time. There are not that many variations. Windows are similar but it depends on the project how I treat them.Kitchens I build from modules with additional unique pieces.

Finally you got the landscape which Unreal is very good in creating.

I think datasmith is great as a transition tool for people who want quick results. But there is a significant amount of fixing to do in Unreal for each scene and you can’t really reuse any of the work done in Unreal. It has to be repeated in the next project even if it is exactly the same model brought in with datasmith. It is only the 2nd best way to use datasmith in my opinion.

I realise that it can be very difficult to master Unreal but so is Blender and 3ds max. I was talking about creating the entire scene in 3ds max or CAD and then use datasmith to go back and forth between Unreal and max. This workflow seems to be only the 2nd best choice to me.

Oh, okay, we are on the same page then and I think I agree with almost everything you just wrote. Thanks for chatting on on it. :slight_smile:

I am looking to do another computer build since my desktop is getting a little long in the tooth. I just built a laptop using the 2080 and the 9th generation I7 and the light build crushed both my offic e and home workstations. I am thinking of the Ryzen for my processor and the 2060 or 2080 for my video card. Have you been okay with the processor you have? I know it is good fro rendering but do you see any lag when working on more of the 3D modelling stuff or other non-rendering tasks?

I bough this processor to use with Corona since it’s CPU based. so far It is doing great. Usually when I’m rendering in 3ds Max I’m also editing lots of images in photoshop with good performance (after changing the priority of 3dsmax to low). Also, I usually have tons of chrome windows and tabs, youtube,… open at the same time. No complains at all.
Be aware it will need a watercooler, I’m using a corsair H 115i to cool it and also, I recently upgraded the amount of memory from 32 to 64gb since photoshop also consumes a lot of RAM.

A great post. I’ve been enjoying just playing around with the new realtime ray tracing capabilities - and the improvements to Datasmith. But, if I’m honest, Epic has a long way to go before I’d ever consider adding Unreal to my toolbox. The main problem - aside from the still dodgy ray tracing items you’ve mentioned - is that the whole thing is just still too finicky and cumbersome. Just looking at material creating and editing is a classic example: it’s far from intuitive and the slate style interface always ends up looking like the diagram for a nuclear submarine - even for a basic white paint material, LOL.

I have a choice: Fiddle endlessly with a scene in Unreal in which - maybe at some point - I can get a decent looking image…or jump over to Enscape and get an absolutely amazing looking experience pretty much instantly - with full RTX implementation. No light probes necessary, heh. No ongoing problems with ray traced refractions. It just works. I can save out as a still image, rendered sequence, panorama or realtime VR. There isn’t really a choice. I may not be able to animate a door opening or have steam rising from my coffee just yet, but the total ease of use and utterly fantastic image quality more than makes up for that… Animation is ‘nice to have’. Fast, incredible quality is a ‘need to have’.

My only hope is that Twinmotion - now part of the Unreal family - full adopts realtime ray tracing in the near future. It’s a beautiful interface - so easy and smart to use. Another thing to keep an eye on is that Chaos is about to release their realtime Lavina player for Vray.

Lastly, the promo videos and webinars for architecture in Unreal aren’t doing anybody any favors. The amount of hype is enormous, and barely matches the reality of actually working in the platform. In a recent webinar with NVIDIA and Unreal, the main presenter was from an architecture firm. The examples looked pretty good…until it was mentioned that they’ve totally built an enormous series of custom scripts to automate a ton of stuff you’d normally still have to do manually - and the script wasn’t going to be made available. Oy vey.

1 Like

unreal engine…world class results…3rd world interface. its a love hate thing. lol. jesus h. world outliner, what a piece of…

imo the best thing epic could do for 4.26 (after the next release) is stop adding new ‘beta’ test features and fix all the ridiculous hoops everyone has to jump through to do the most basic of things. i would love ue to be a user friendly well behaved app insted of the supermodel diva ■■■■■ it is today

2 Likes

I get why you are making this comment and I agree with the intent of it. However, in the archviz world, I use other software like Revit, Unity and some other things where they have their own versions of asset management in the form of outliner and I can’t say it is much better.

i tried revit for a few hours and quicky came to the conclusion that it was the worst god **** awful spawn of satan ever to roam the earth…much like all autodesk software…and i am fluent in both inventor and autocad despite them both being total and utter *****. i hate autodesk with passion. infact i have recently had to use autocad for the first timr in 15 years, and it is still doing the same dumb shiit that makes you roll your eyes as it was doing 15 years ago, and not alot has changed since then with the app.

quite convinced that there is a team at autodesk that deliberately schemes to make their software as user unfriendly as possible, just for fun… but probably to sell training courses and subscription lockins.

and i LOVE epic for their passion and devotion to providing tools like ue for us for free, and i would love for their ui to be part of humanity.

for instance … in every other application ive ever used, when i press ctrl+ z to undo, the last action is undone and the ‘thing’ remains selected so i can complete whatevr it was that i was trying to do…this has been universal for the 30 years ive been using software and is what everyone wants and expects. except epic have decide that i want to also deselect the thing and force me to hunt for it again in a world outliner that someone has decided HAS to be alphabetically sorted regardless of what i want. i literally have to make folders numbered 010, 020, 030 to keep myself organised like i was programming basic on my c64 in 1985. who asked for forced alphabetical sorting and decidee this was theonly way wo would operate?

ffs. if i was a deceloper, (and i kind of am at times), i would evaluate other software to establish what was the most efficient and user friendly method of achieving a common goal, and ue really isnt it, as much as i love the result.

and what really bugs me is that this and 1000 other silly annoyances are trivial things that could be fixed with almost no effort for everones benefit. its obvious that ue is at the cutting edge of tech so why is it so difficult for them to fix a fxxxxing listbox sort algo? (and i have read other comments and requests for ue going back years, so im not the only one who despises the ui)

BUT i have to say, ue is awesome for what it can do, and im prepared to stick with it in anticipation of where we will be in a couple of years from now.

Interesting thread, My Thought using UE 4.24 for ArchViz is - UE is not reliable for archviz professional big scale projects, for small scale may be it can be good but big scale projects you cant rely on ue4.
I have been trying to test medium scale interior project but couldn’t succeed i was wasting most of the time finding solution for errors one after another, imagine if it was a real world paid project with deadline, i am sure client would have been cancelled the project eventually because of time and he’d have lost trust in me for any future projects.

Imagine if you try to work on exterior big scale project where will you end up.

The projects what we see on youtube with realistic quality either its small project done in free time or it is done by big studio with big technical team.

2 Likes

Do you mean specifically UE 4.24 or using Unreal Engine in general with ArchViz? If you are referring to 4.24 specifically, I can understand that since this release does nto seem like a great release for ArchViz users. Someone just posted somewhere that the compiling issues will be fixed in 4.24.2 and then I saw it will be 4.25. I am not doing any projects in 4.24 as I just do not trust it for the very concerns you mentioned. I cannot risk losing a client with delays trying to troubleshoot.

As for using UE in ArchViz, I have been getting better at establishing a workflow that leads to some pretty predictable results but these have not been large scenes. I am about to try again with a new scene (interiors model) that will be a bit more robust. I also use Enscape which is a real time engine specifically for archviz that is pretty hard to beat in regards to ease of use. However, I want to go much more custom so I keep playing with Unreal. If it were not for Datasmith, I would have had to probably abandon this effort awhile ago.

The thing about NOT using realtime renderers is you get into significant amounts of rendering time to generate fixed renderings and even more time to generate any sort of walkthrough. I am predicting these offline renderers are going to be less common as realtime technology keeps improving.

One way to speed up data transfer - especially for large projects - is to use the newly added Data prep editor where you can define what and how things are imported from other packages…yes its still in beta and needs to be activated via the plugin to work but it could make the difference if you’re on a deadline…once you set it up you can just re-import new files with ease…

Unreal offers perhaps too many options but if you focus just on what’s important to get the job done its not that bad…I used to work in vfx standard rendering workflows before switching to realtime…and yes it took me a while to reestablish my orientation - the same workflows do not apply…however the basics do - good modeling is still paramount, attention to texturing and good lighting will get you very high results in Unreal…

In any case if anyone’s interested I have a couple of tutorials on youtube about raytracing and beveling edges - as someone mentioned was an issue for him…

I thought I was going crazy until I ready this thread. It’s been so useful. Thanks.
We recently ran a production job using UE for both realtime presentation and high res stills + animation. The client was expecting “same as Vray” results. UE4 could not deliver the same level of detail and control that most clients we work with require. As we pushed for that goal, we hit a hard ceiling of technical limitations. The render time benefit went out of the window.
I think we really forced the wrong solution in this case (live and learn I suppose).
What I would love to see is a unified content source that both DCC and UE4 can read from. A live link of sorts. Make a model edit and it syncs to both. USD will hopefully feature here.
I’m looking forward to seeing more development in the future. Nice to know I’m not “doing it wrong” though :heart_eyes:

The epic team could have done that via Blender when they made their blender plugin that doesn’t work.

Instead of listening to the feedback of others who had already made a plugin directing them to simply allow the engine to read a .blend file (which is not something in the scope of what plugin makers can do), they continued to use the FBX pipeline.

Now, this isn’t to say it would have been useful to any architect’s workflow. But it would have been a step in the correct direction by allowing at least 1 program to have its file natively opened within unreal.

As far as ‘why blender’…
Well that’s because it’s open source. There’s no licensing to worry about.
It was also around the time they donated over 1 mil to the blender foundation, so no one would have ever complained.

Instead, you get to keep old pipelines that waste your time, and a sub-par rendering pipeline that’s full of mistakes when it comes to physical rendering.

Though for achviz you may be better off using the raytrace nvidia branch and taking still.
If you spend 2 or 3 days working it has potential to look good when you take a high res screenshot.