Witch Atelier Interior Environment
This page is currently under development! It is being used to document my process and progress with this scene, from start to finish. The current state of the project and next steps will always be at the top, with the full process documented chronologically below.
Current state of project
Currently, most props have final or close-to-final highpoly geometry, though some need polish.
Next steps are to complete the remaining prop geometry, then to finalize and revise the architecture geometry.
Next steps are to complete the remaining prop geometry, then to finalize and revise the architecture geometry.
The Process So Far
Project Goals and Core Concepts
As with any artistic project, the first step to success is to define personal goals. Since this scene is a collaborative effort between me and a foliage artist, these are goals that must be shared between the two of us.
Goals for the project |
|
First steps--Concept and Reference Gathering
With goals defined, it was time to gather reference. The images below are indicative of the current state of my reference board--it has grown considerably since the start of the project.
In order to gather appropriate reference, I scoured physical reference books, pinterest boards, and wikipedia pages for primarily photos or art book content. I found some real gems, including an incredible art nouveau dome that became the inspiration for the quarter-dome in the piece.
In order to gather appropriate reference, I scoured physical reference books, pinterest boards, and wikipedia pages for primarily photos or art book content. I found some real gems, including an incredible art nouveau dome that became the inspiration for the quarter-dome in the piece.

Once I was satisfied with the reference I had gathered, I began sketching thumbnails of a number of possible spaces. All had their merits, but in the end, I picked the most complex option, which is the top left image. I felt it would be the most unique and visually interesting, and the difficult modeling challenges could emphasize my proficiency with complex architecture and forms.
From here, I could have fleshed out the sketched concept to get a better idea of what the final scene would look like. However, i decided that gathering more reference and experimenting more after blocking it out would be more beneficial.
Initial Scene Blockouts and Scale Tests

My initial mesh blockout gave me some good information--mainly that things were currently a little too big. The scale overwhelmed the guy, and the door was a good reference for how much bigger the space was than it should be. It was also somewhat difficult to frame shots with the space so big.

My revised scale test shrank the overall scale of the room a bit. It needed to still be a bit large and grand, but not cavernous like the original blockout. To better get a feel for composition and scale, I also began blocking out some more elements such as a cauldron, plant pots, and cushions. In addition, I had gotten the feedback that a reading nook in the arch cutout would be fun and interesting, so I added that in along with a ladder. The scene was starting to come together a bit more.
Initial Mesh Blockouts, Layout, and Lighting
Now began the tough part. I needed to start properly blocking in the architecture and primary elements like the desk and shelving. Using extensive reference, I began to develop core scene language, taking many cues from art nouveau and 18th century european furniture. Architecture primarily came from art nouveau sources, but toned down to not overwhelm the scene with austerity and maximalism. The built-in shelving was inspired by secretary desks and kitchen cabinet hutches.
I made sketches at this time, roughly trying to figure out designs that would work. These designs would change radically as the scene developed, but they were very informative at the time.
I made sketches at this time, roughly trying to figure out designs that would work. These designs would change radically as the scene developed, but they were very informative at the time.
Using my very rough sketches, I began blocking in these forms and bringing them into unreal for scale testing.
The desk I immediately realized would need revisiting, tossing out my original design entirely. Other parts of the scene retained more of their original elements, but the desk would eventually be re-designed to incorporate the shape language of other pieces.
Still, the scene was beginning to come together, with aspects like the dome and arches bringing a lot of visual definition to the scene.
The desk I immediately realized would need revisiting, tossing out my original design entirely. Other parts of the scene retained more of their original elements, but the desk would eventually be re-designed to incorporate the shape language of other pieces.
Still, the scene was beginning to come together, with aspects like the dome and arches bringing a lot of visual definition to the scene.
Up until this point, all the meshes in the scene had been imported together as one mesh for the sake of quick iteration. It was far simpler to define the overall scene when I didn't have to worry about re-exporting every single individual mesh every time I made major changes. But now, with the core aspects of the scene locked in, it was time to re-setup the scene using individual meshes.
At this time, I also took the opportunity to block out some new props, such as a book carousel, cabinet, and side table.
Once everything was set up in a fresh unreal scene, I took a very first pass at rudimentary lighting. I made sure to place some cube meshes behind the walls to prevent light bleed, put in a placeholder frosted glass material, and began playing with my stationary light. I wasn't trying to achieve anything ground-breaking, just enough to start from.
At this time, I also took the opportunity to block out some new props, such as a book carousel, cabinet, and side table.
Once everything was set up in a fresh unreal scene, I took a very first pass at rudimentary lighting. I made sure to place some cube meshes behind the walls to prevent light bleed, put in a placeholder frosted glass material, and began playing with my stationary light. I wasn't trying to achieve anything ground-breaking, just enough to start from.
Mesh Development--Architecture
Now that base forms were in and looking good, it was time to start developing more final geometry. I began with the semicircle arches, making use of opensubdiv and shell modifiers to achieve the interesting shapes on the corners and up top, while keeping everything nondestructive for future iteration.
Next I cleaned up the dome blockout geometry, which had been fairly messy. Since the dome is not a perfect quarter-circle, and is instead a soft oval, the circle math was a bit of a challenge to deal with. However, I managed just fine, and the result is mostly clean, with the understanding that the top would still need some iteration in the future.
Finally, I went back to tackle the support arches. These were a real challenge, and I ended up completely redesigning them from my original plans. I made the filligree more significant and detailed, and added flower lamp blockout geometry to iterate more on in the future. It was here that the floral theme began to solidify as shape language, which would be populated to other aspects of the scene.
Column capital
In order to complete the arch, I knew that I wanted to put in a detailed column capital. Gathering various reference, I drafted a rough concept in CSP and began blocking it out in max. It changed as I modelled, becoming more detailed and better suited to the surrounding geometry and space overall.
Once I was happy with the blockout, I brought it into Zbrush and worked to strengthen the edge definition, blend the flower forms more strongly into the base geo, and to add an extra degree of detail to the leaves and stems.
I didn't want to have to retopo immediately, but the weight of the zbrush mesh forced my hand, and I was forced to make a rough lowpoly so that it wouldn't crash my Unreal or max scene. I did a quick UV and bake, bringing the AO and normals into Unreal and setting up a first-pass material on it. It still needed another pass, but since I planned to include it on a multitexture sheet in the future, it would need to be revisited anyways.
Once I was happy with the blockout, I brought it into Zbrush and worked to strengthen the edge definition, blend the flower forms more strongly into the base geo, and to add an extra degree of detail to the leaves and stems.
I didn't want to have to retopo immediately, but the weight of the zbrush mesh forced my hand, and I was forced to make a rough lowpoly so that it wouldn't crash my Unreal or max scene. I did a quick UV and bake, bringing the AO and normals into Unreal and setting up a first-pass material on it. It still needed another pass, but since I planned to include it on a multitexture sheet in the future, it would need to be revisited anyways.
Temporary Glass Shader
Needing a break before going back to modelling after the intensity of the column capital, I took some time to experiment with getting a frosted glass material working. After some investigating, I got the essential principles I would need down and put in a temporary material. This would help to better define lighting in the scene, and gave me some peace of mind that a material like this would be possible for me to achieve.
Pallete Formation and Initial Color Blocking
Around the same time I started in on the glass material for the dome, I also began nailing down a color pallete for the scene. To do this, I analyzed 25 images which had the correct mood and tone for what I wanted to achieve. I wanted bright colors, accented by wood and vegetation. By combing through these images and extracting the core colors from each, I got a better feel for the pallete that I would likely target, and made some swatches to sample from when color-picking.
Next, I set up a multi-material in 3Ds Max to apply to the architecture mesh. In Unreal, I could then attach placeholder materials to these slots and experiment freely. I wasn't entirely happy with how it was looking, but the setup work alone was useful and necessary.
Furniture Modeling--Shelf
Since the architecture of the scene was now more or less figured out, It was time to begin modelling the other large elements. The biggest one I wanted to start with was the in-wall shelving, which was a complex mesh that took a number of days to complete.

I began with the basic forms, of course, then added support deco pieces on the top as well as swirled column decorations to echo the support column geometry.
Then, I made a drawer knob that would end up being used many times throughout the scene, and which I planned to eventually make a baked version of.
After that, it was the cabinet filigree, which became the base for the drawers. The secretary desk-style cubby holes and drawers came after that, which were very fun to model.
The tileable floral design featured on the left was used as a moroccan-style grate on the bottom under-desk, and I also planned to possibly use it elsewhere in the scene.
Once the shelves were feeling closer to finished, I brought them back into unreal and made some scaling alterations to better fit the space, adjusting the desk proportions.
Color and Lighting Revisit
Now that so much of the scene was coming together, it was time to experiment a bit more with materials. I properly UV'd some of the architecture portions, and started playing with placeholder materials to see what I liked and what I didn't. However, I couldn't quite get something I liked.
So I dragged the scene into CSP, and began to do a paintover, not only experimenting with the colors but also with lighting. When I had something I liked, I began to incorporate some of those changes into the scene, adding some geometry to the walls and adding some more materials to the multi-material. The dome was going to require its own solution, so I largely left that alone for the time being, just adding a bit more geo on the cap.
Small Props Modelling
Now that things were really progressing, I decided to take some time to model the smaller props I'd be using to fill out the shelves, desk, and other aspects of the scene. This was really fun, even if it took a little bit.
As I worked on them, I began placing them and moving things around to my taste. The scene was really starting to look like something!
Large Furniture Modelling -- Side table, Cabinet, Desk
With the small props finished, I began work on the last of the larger wooden furniture. I was able to reuse a lot of geometry from the shelves, with plans to create panel multitextures for the aspects shared between pieces. Sharing the detail work saved me time, while also giving the pieces the appearance of being in a set.
Put together in the scene, I was happy to see how coherent everything looked together.
Medium Props Modelling -- Chair, Bench, Table, Ladder
Next up was the chair, which was inspired by antique piano stools and art nouveau dining chairs.
After that was the bench, which was quite challenging to block in given its curved shape. I left some detail to finish up later, but was happy with how it started to balance out the scene's weight.
The table was next on the list, and was definitely a fun challenge. The legs, especially, required a good amount of reference and redesign.
Finally the ladder, which I was pretty happy with.
Scene Re-assessment
While I was hard at work on my own props, my foliage artist friend began work blocking out the pots she will use to contain her plants, as well as a hellebore. With the pots and hellebore placed in the scene, things were looking closer than ever to final.
At this point, the scene's geometry is just missing a few last props, and I'll soon need to move on from purely modelling to material and shader work. Since I have plans for many aspects of this scene to share textures, this means a certain amount of planning. I created a kanban board using miro to plan next steps, as well as to plan my multi textures.
Stained Glass
Initial Design
With most of the geometry determined, it was time to figure out the design for the stained glass dome.
I began with drafting a rough concept of what it would look like, then UV'd the dome and did a quick sketch-over in substance painter. From there, I brought it into CSP to create a first pass of the texture.
Things were a bit rocky. Putting the shader aside, the design lacked cohesion and the first pass of the upper dome section was far too busy to be readable.
I began with drafting a rough concept of what it would look like, then UV'd the dome and did a quick sketch-over in substance painter. From there, I brought it into CSP to create a first pass of the texture.
Things were a bit rocky. Putting the shader aside, the design lacked cohesion and the first pass of the upper dome section was far too busy to be readable.
Edits and Adjustments

I wrote myself some notes on how to improve the design, and quickly went back to fix it up and get a better grasp of the changes.
Unfortunately, the way I UV'd the dome led to too much stretching for my initial plan of how I was going to do the upper dome. This led to a big headache, and even after the nightmare was over I still have stretching on the near corner. However, due to the position of the camera, i decided to put aside the stretching for now, since it was fairly minor and I already had quite a migraine. However, after all was said and done, I had something I was happy with for now, with the knowledge that I would have to do some significant revisions of the textures later in the process. |
This is where I decided to leave it for now, with plans to revisit it later in the texturing process.
Remaining Props Modelling
Taking a break from the dome, I decided to tackle the rest of the minor props that still needed modelling.
First up was the cylindrical bookshelf carousel, referenced from a collection of antique similar carousels.
First up was the cylindrical bookshelf carousel, referenced from a collection of antique similar carousels.
Next was the wall shelf, which had some very fun shapes I referenced from some ghibli art.
Then it was the shelf dedicated to hanging herbs, which has some fun hooks and shape language borrowed from other pieces in the scene.
Marvelous Designer Work
The last collection of props I needed were all fabric-based assets, so I decided to hop into Marvelous for the first time.
This was a challenge-- Marvelous is a fashion program first and a modeling program second, Learning it took a good amount of time, but by the end of a number of helpful tutorials I finally got the assets I needed out of it.
This was a challenge-- Marvelous is a fashion program first and a modeling program second, Learning it took a good amount of time, but by the end of a number of helpful tutorials I finally got the assets I needed out of it.
The blankets were first, because it was the easiest to get started with. I imported them into Unreal and enabled nanite on them for now, and it might stay like that since the UVs are already done from Marvelous.
Next I tackled a much more difficult mesh--a button cushion. At first I tried to simulate it myself without any guidance, but ran into enough problems that I gave up and followed a tutorial. The tutorial taught me a ton of stuff-- such as how to use the sewing tools, the differences between the various simulators, and a bit more about how to convince Marvelous to do what you need it to.
Lastly was the square cushions, which were far easier than the button cushions. The fringe I did in max, with an array modifier on a spline.
Architecture Polish
Next up I revisited the parts of the architecture that were still in blockout. I reused some of the blockout geometry for the hellebore carvings to make the arch cap and lamp. I also modelled shelving into the cubby space and propped it out, including some placeholder lighting.
Then, I hooked up a spline blueprint to generate the wires and hand-placed fairy lights with a bit of emissive on it. I hope to use this same spline blueprint later in tertiary propping.
Then, I hooked up a spline blueprint to generate the wires and hand-placed fairy lights with a bit of emissive on it. I hope to use this same spline blueprint later in tertiary propping.
Door and Stained Glass
The most complex bit of geometry left to model is the door, which I did a sketch over in CSP. Next, I modelled the parts that needed modelling, and checked scale and presentation in Unreal before moving on to UVing the stained glass.
I added leaves to fill out the design more, normalized some of the strokes, cleaned up some things, and then finished up the color and faked AO. I generated a quick and dirty heightmap and opacity mask, and was very pleased with the results. Perhaps colors will continue to change, but the design for the door is likely final. |
Plants and Full Scene Review
The plants are the responsibility of Caitlyn Kenney, my dear friend and foliage artist from Ubisoft. Her work is really bringing the scene to life, and making me very excited to finish the last of the modelling polish I need to do and finally begin texturing.