Interior room EXR texture maps for architectural visualization

[Giorgi Nanava]

Hello,

I'm trying to figure out  how to render an interior so I can create a similar map.
I need it to create EXR interior parallax map. 
Is there any tutorial I can follow or no hope here and I need to figure out it myself?

Thanks!

[James Vella]

Assuming you use vray, or if you dont check your render engine for cubic map rendering.

https://medium.com/updates-from-vizor/how-to-make-stereo-cubemap-architectural-renders-with-v-ray-a9c77c8d1988

 

Then just save as .exr format

[Giorgi Nanava]

Thanks! Yes, I use vray!

I checked the article, and it is about 360 stereo cube map, but I think it is not exactly what I'm looking for - What I need is an orthographic views of each wall, floor and ceiling. 

I try to dive into this tutorial and figure out may be I can tweak it enough to get the result I need, but not sure yet. 

Thanks for your link!

PS. I think that I'll need to render manually all views and combine them in an EXR format? Lot of manual work but don't see better options. 

Edited November 5, 2022 by Giorgi Nanava

[James Vella]

If you want orthographic views use "orthographic projection" in the standard camera

Then make sure you use camera clipping to cut the walls from the closest part of the camera

 

Regarding the EXR/combine, this is not a lot of work, 5 minutes in Photoshop and you are done. Copy paste, align, repeat, save.

 

Btw, not sure what software you are using to create a parallax map, but most likely you will need the zdepth data for each image (which works better with perspective view not orthographic, but anyway). Again dont know your specific situation since your description is so short but thats my 2c.

 

Edited November 5, 2022 by James Vella

[Giorgi Nanava]

Yes, that's exactly what I'm gonna do, and you are right about lot of work, it is not. 

I'll be using Sketchup, for the project - I don't think I need zdepth data, I need to create parallax for interiors like in the attached image. Here all I need is orthographic images of all walls, celling and floor - and cutout renders of furniture, all saved in EXR and organized in a 3X3 Squares in photoshop, like in the image in my first post. 

And here is how to use the EXRs in Sketchup - looks simple and very effective. 

I can be missing some steps in between but so far that is what I found out - will come back with my results when ready. 

  

[James Vella]

I dont use sketchup so cant help you with the program specifics. 

 

Well usually parallax is 1 image on a plane with depth data, as the camera moves it projects the depth depending on the angle of the camera, thus the need for the depth data. 

If you plan on projecting each image onto the floor, ceiling, walls and then doing cutouts for the furniture....that sounds like a lot of work, and a very old way of doing things and the shadows from the furniture will be difficult to line up. Not sure the intended output but cant you just do the foreshortening in after effects or whatever comp software?

 

I would need more to know more about what you are doing, everything is quite vague at this point. I dont feel confident I would give you the right advice as the goal post seems to keep moving. Give us a proper description of what the outcome you are looking for, do you have a video or something similar as reference? What files do you need to deliver, is it a video for a client, are you in a pipeline and need to give this to someone else etc?

Edited November 5, 2022 by James Vella

[Giorgi Nanava]

Sorry for not being specific, in fact it is very simple, I'm Architect and need to create great exterior renderings for my projects. For the exterior rendering to be great it needs to have some interiors that you could see through windows. Modeling everything takes time and lot of PC resources - so there is several options that help with that: ShaderBox, wParallax and Evermotion Archmaterials  Vlo. 4 - you can use these maps in 3DsMax (V-ray) as an emissive material and instead of modeling whole room you have a rectangle with material that looks like a furnished room. So these materials are available and I can buy them, but I'm trying to learn how to make them myself, so I can create my own custom interiors. That will give a great flexibility. 

These materials also can be used in Sketchup but with a little workaround - but right now it does not matter which software I'll use. 

Here is the link to the Dropbox folder where you can find the files I'm talking about - they are from Shaderbox. 

https://www.dropbox.com/s/geqio6gisumk6x5/Starter_Kit_Scene_Files.7z?dl=0

link to video: 

 

Edited November 5, 2022 by Giorgi Nanava

[Giorgi Nanava]

No, I was wrong, the renderings must not be orthogonal. Don't know why I though so. Just figuring it out )))

Edited November 5, 2022 by Giorgi Nanava

[James Vella]

Ok that makes sense. Its a rudimentary way of doing parallax with ShadersBox but simple enough to follow. 

1. Create your interior room, set your cameras up to capture the walls/floor/ceiling and one for the furniture/curtains. Cameras will need to use clipping since you will be behind the walls.

2. Render each shot save to an image/exr, each of these will be a separate image combined into 1 at the end. Furniture/curtain renders will need an alpha so they are a 'cut-out' (no background). They will need to cast shadows onto a matte floor (see through floor, shadows visible on alpha/matte).

3. Bring these into Photoshop and lay them out as per your reference. It actually doesn't matter how you lay them out but what does matter is that every image you do from here on will be consistent with the first layout. This is the 'magic' here, so when you have 10-20 rooms rendered you can just swap out the texture with the next one and they will line up with each other (floor is always floor, back wall is always back wall, if you change the order in Photoshop then your back wall will swap out with the ceiling for example since its just a texture atlas basically).

4. Follow the steps below in the video for the technical stuff. 

 

By the way from the looks of it exr is not really required, that's up to you. However its still probably a good idea since you can embed more data - such as light emissions from lamps/ceiling lights for extra control, alphas/opacity, material ID selection, depth etc. I wouldn't worry about this at first, get 2 images working with parallax then you can play with this to see what else you want to control, then build this into your workflow. Only you know what you will want to adjust later so in the early stages its best to build this into the exr. 

 

Regarding the cameras, whatever focal length you choose for the front view (furniture/back wall, curtain) keep that consistent since you want your shadows/perspective/etc to line up. Keep this as a template for all your rooms you render so you have some consistency or things might start to look off from room to room. The floor/ceiling in this situation will need to be rendered with quite a long lens and parallel to the surface, just play with it a bit til it looks ok.

 

I think the OSL/3dsmax (wParallax) way looks better imo, its using parallax on a single plane as I mentioned previously, much easier to setup in a quick fashion since you only need 1 render (front view, no alpha required, no need to uv each wall/floor/, no need to place furniture/curtains, no need for additional camera setups etc). This is using the depth built into the exr image I was referring to, but if you have to do it this way then that's what I would do.

 

Gotchas: Another important thing to be aware of is that if you go with the ShadersBox way of doing it you are limited to what you can change later, for example if your template has 7 renders (floor/walls/furniture/curtains), if you want to have another layer later then you need to build this into your template else it will cause you headaches when swapping out the rooms for a different one, this also has other down hill headaches. Consistency is key in this workflow. Its also a hell of a lot more work since more layers/cameras/renders means far more things can go wrong or consume your time. If its possible I would certainly find a way to go with the wParallax - not sure if this viable for Sketchup but might be worth your time investigating. 

 

 

Edited November 6, 2022 by James Vella

[Giorgi Nanava]

Hey thanks for your instruction and tips, especially about the camera's focal length - that is really important! 

Regarding the usage of exr shaders, I'll be using SketchUp as it is more convenient for me, the overall workflow of modeling and rendering is much simpler then that in 3DsMax, and I'm ready to trade the better OSL shaders for the manual workaround in SketchUp. May be once SuP will also be supporting OSL language and then life will get really good : )))

But what do you mean when saying that I'll need only one render to create the OSL map in 3DsMax? Do not you have to create different views and stitch them up so you have the usual 3x3 exr map? I'm ok creating exr maps in 3DsMax if that is faster. 

Thanks again for your help! 


 

[James Vella]

From the example they showed in the video, its just one image on a plane, no need to render additional views or do any wall/floor mapping or cutouts for furniture. So no you dont need a 3x3 map or whatever, its just a render from the front view of the room.

 

We used to have a similar technique with an in-house render engine about 10 years ago that did parallax mapping this way (before OSL). Basically we just render the room from the front angle, save the texture with a depth map and put it on a plane. Put the plane behind the windows and its done. As the camera moves it changes the perspective of the image using the depth data. It looks like the same technique as the video you posted earlier - you can see the first 10 seconds when the camera is rotating look behind the render window you can see its just a single plane, not a box.

 

There are lots of benefits to this to this method:

- You render once, its easy to update and change things for that shot without re-layering etc, you don't need to ever open photoshop.

- You can re-use the same scene and to make a different room without ever creating a new camera or changing any settings.

- No furniture cut-outs, mattes, alphas etc.

- No uv mapping required.

- Far less things can go wrong, like I mentioned earlier when you are dealing with a 3x3 layout and 7 different renders per window image you have more variables to deal with.

- No R&D time trying to work out camera focal lengths and consistency across the board.

- Far less work - 7 renders per image vs 1 - unless you want some with window blinds and some without, but instead of putting that in the texture I would personally put this as a separate plane/geometry and use some parametric placement tools to put this in front of the parallax plane. I don't know the specifics of how they did it in this particular case but my hunch is you would need a little extra work with layering things in the OSL so for me I would go with the quicker method of the two.

- Distortion is universal - meaning when you rotate to an extreme angle either method is going to look distorted to some degree however when you have cutout furniture you are going to get some foreshortening issues. Most clients probably wont notice at first however once you see it, it will be the pink elephant in the room.

 

There's probably other things I'm not thinking of since its been awhile but that's the gist of it. I'm sure anyone whos currently using this technique may be able to jump in and fill in the gaps or correct me if I'm wrong somewhere.

Edited November 6, 2022 by James Vella

[Giorgi Nanava]

From the files they provide I know that the EXR files are organized just like in ShaderBox's version 3X3, you just need OSL script for them so they work together. Because Sketchup does not support OSL you have to build boxes and project textures to create similar effect. that is the only difference between 3DsMax and Sketchup. 

Thanks for your help!

 

[Giorgi Nanava]

More I learn about OSL maps, I see how much more control you can have over them in 3DsMax - you can create random variations from a single map -  that is cool. 

[Giorgi Nanava]

And here I found YouTube tutorial how to render image in one go.
I don't understand a single word the guy is saying, it is in Arabian, but I think video alone is enough - I can figure that out.
And another Video in Spanish. No videos found in English - so interesting. 

 

 

[James Vella]

Ah yes you are right it seems that the OSL uses the same layout method on a plane. So I suppose in this case the position of each render in the 3x3 should be as the specification. Looks like you have enough information to go from now yeah?