The shoe

The shoe was the first modeled object from the entire project. It was also the first object I modeled using the box modeling technique. With this object I learned the basics of this technique, so it took me one week to model it. The fact that I had to study the real shoe, by observing it in the videos and in the photos I had, contributed to the difficulty of modeling it.

The rose-angel

Modeling this creature was quite interesting. As one may notice throughout my WIP images, I had many versions of its wings. Being the center of the image and the subject of the entire project, I was really picky about it. Based on suggestions from different friends, I remodeled the wings several times. At the end, I decided to stop reaching perfection and decided to be content of the sculpted wings one can see in the final render.

The final wings were textured with Bodypaint.

Birds

These birds are really far in the background, thus the simplicity of their modeling is obvious. The textures were mapped on the objects in a very weird way. At that given time, I did not have knowledge about how to do proper UV maps. So, for this bird I manually mapped multiple materials on the object using the predefined projection modes, huh.

The grasshopper

Due to lack of rigging knowledge, this creature was modeled exactly in the desired pose. The modeling itself is not of very high quality. I intended to keep it simple, because it will be in the background of the image. I haven't done any studies about anatomy. This would have been required to make a correctly detailed grasshopper.

To texture the creature, I used for the first time Bodypaint and I learned to work with UV maps. The painted textures are at a very high resolution, over 5000 pixels. After the pleasing experience of painting in 3D, all the objects I've done after this one, were painted this way. At first, dealing with UV maps wasn't easy, but after a while I got used to them.

Butterflies

The butterflies were probably the most enjoying and creative thing to do for the entire image. The modeling of these butterflies was, obviously simple, but the texturing wasn't so simple. I learned to use Photoshop even further. For the first time I used on a daily basis the Paths tool. All the designs were manually drawn in it and colored afterwards. Each wing texture has a resolution of 1024 x 768 or higher. In the WIP images, you'll see that in 3D, I made their wings look like fabric. The designs of the wings are based on various sketches and drawings of butterflies I found on the Internet.

Flowers

Without any doubt, doing the flowers was also a fun thing. I enjoyed the techniques I used for modeling and texturing them. I've learned to use xFrog extensively after trying various ways to make plants. Most of the textures are shaders, with few exceptions, which are done in Photoshop.

Trees and vegetation

All the trees and the vegetation was done using xFrog. For all of the plants, I created different shaders to texture them. No raster images were used. Of course, it was a true pleasure toying with the shaders and trying to achieve good results. On the boring side of things... I had to manually distribute them, one by one. And to the extreme, I had to manually move the grass blades intersecting with other objects. I found no automatic way to fix this.

Scene management and related gimmicks

As always, when I do a 3D scene, I start by modeling the objects and by doing the composition of the image. For each part of the image I did the same.

The scene was really hard to work with, because of the very high amount of polygons it had. Texturing and lighting it all in a single scene would have been nothing but a hurdle. So, for the first time, I used the technique of rendering in layers. I separated, based on depth, the big 3D scene into smaller ones. In the end, I had 6 scenes. Having a big 3D scene split like this, also increases productivity and overall quality of your final product by allowing your to better concentrate on the smaller parts of it.

The layers of the image:

• The forest
• The two trees
• Vegetation
• Flowers, grass and the grasshopper.
• The shoe, rose-angel, water, petals and the water lilies.
• The arch and the butterflies.

Everything went smooth. All layers were rendered as HDR images. The only show-stopper, technically speaking, was the water I had in the 5th scene. Obviously, when rendering... nothing was being reflected in it. So, after thinking about a solution, I decided to render the reflections for each layer and then compose the images into Photoshop. The reflections were all rendered as HDRIs at a resolution of the final render: 2756 x 3528. The final reflection map, in .PSD format, had 660 mb. This image was projected frontally to the camera view.

I made the water very dark and blurry on purpose. The image was already chock-full of elements and a bit too cluttered. So in this way I avoided making it look more noisy or cluttered. Having the water purely reflecting everything would have looked bad.

Final render

All layers were rendered with Anti-Aliasing.

I've rendered the mask of the 6 layers and different depth-maps. These were my main "tools" for post-processing the entire image. As you may have seen, all the variations of this image were obtained in this way. I enjoyed doing them. I worked in RGB 8 bits mode, but dithering errors were too obvious. I had to "render" the images in 16 bits mode. I was not able to work constantly in 16 bits due to PC memory limitations. The Photoshop file for the final render with all the post-processing has 600 mb.

Scene statistics - part 1:

• Objects: 68 243
• Polygons: 2 917 023
• Lights: 194
• Materials: 179
• Project size: 5,48 Gb

Technologies and techniques used:

• SubPolygonal Displacement
• Ambient occlusion
• SubSurface Scattering
• Hair module from C4D
• Volumetrics
• Box modeling
• 3D Sculpting

In my initial plans, I was thinking of using Global Illumination and Motion Blur. Due to lack of CPU horse power these plans were canceled.

Approximate render times:

• with an Athlon XP 1800+, 1 Gb RAM:
  • Layer 1 - 10 hours
  • Layer 2 - more than 10 hours
  • Layer 3 - 24 hours
  • Layer 4 - maybe more than 11 hours
• with a Core 2 Duo E6700, 2 Gb RAM:
  • Layer 5 - less than 10 hours
  • Layer 6 - less than 5 hours

The first four layers were rendered in November 2006 in a week or two. The last two renders were done in May 2007.

Attention: The render times aren't exact due to lack of interest in them and due to the very long period of work on the entire project. The resolution of the renders is: 2756 x 3528 pixels.

The work on the right side of the image was started even before completely finishing the left side. I've decided to work this way just to allow myself to think more on how to finish the rose-angel, as I wasn't content with it at that time.

This was my first dark-blooded image. Despite this, it was fun doing it mostly, because I bought my new PC to work on this project.

The building

Modeling the building was quite simple. I only had to let myself inspired by the references from the "À contre-courant" music video of Alizée. It consists entirely of primitives, mostly cubes, nothing else. The interesting aspect of the building is just the texturing. It is all based on shaders and nothing else. For drama, I added some dust planes which are bitmap textures. I used booleans to make cracks and dents.

The hand

I used the box modeling technique to make the hand and then, to improve it, I did some sculpting on it. The texture is all shaders, except the blood which I painted manually with Bodypaint. I finally decided to learn some basics in rigging to help pose my 3D characters and to get more advanced in 3D. This is my first rigged object. It was fun for me to rig it.

Dolls

For the dolls I used the box modeling techniques also and I also rigged them. I used the Cloth dynamics from C4D to relax the clothes on the characters. For the first time I combined dynamics with rigging. Pretty fun. I also used the Hair module. The cloth textures were manually painted in 3D.

The shoe

I already had it made from part one of this project. I wrecked it by using sculpting tools and added details to it in this way. The textures are entirely shaders.

Ground zero

Maybe the hardest thing was to generate the cracks in the ground and to distribute the debris. I had to manually place deformed small cubes around the impact zone. I also used SPD for the cracks in the ground. I added dust around the debris to make the image more interesting.

As you may have noticed, I have a thing for shader based textures. So, the ground is textured in this way too. The not so obvious thing is that I added bump maps for it. The bump map strength had to be high to be visible due to the lack of a high contrast illumination. This is not a bad thing in itself. When I enabled the reflections for the floor, they were all distorted. I did not want them like this. I didn't want to drop the bump map, so I found a solution: add a new plane with a glass like material to get the reflections I want. This has dramatically increased the render time.

The lighting in the scene doesn't use GI. I've done my best at the given time to fake it.

The other objects were done in the usual mode: box modeling. Nothing special.

Scene statistics - part 2:

• Objects: 2 946
• Polygons: 729 636
• Materials: 128
• Lights: 73
• Project size: 3 Gb

Technologies and techniques used:

• SubPolygonal Displacement
• Ambient occlusion
• SubSurface Scattering
• Hair module
• Volumetrics
• Box modeling
• 3D Sculpting
• Rigging
• Cloth Dynamics

The total render time was 45 hours and 36 minutes on the new PC I had: Core 2 Duo E6700, 2 Gb RAM. I've rendered in a different pass the Ambient Occlusion map, which took 8 hours. The resolution for these renders is: 2756 x 3528 pixels.

Post processing the second part was done in a very similar way to part one. I used depth maps and different masks for this.

Thank you for your time. In the following pages I will present various images saved during the project.