You should equalize your textures (remove shadows and highlights) before running them into PixPlant, because it use the luminance of the texture a lot to guess the displacement. If you don’t know how to do it you can check this tutorial on Pawel Filip website. PixPlant has his own equalize texture button, but you have a lot more controls if you do it in Photoshop using Pawel tutorial.

Generating the seamless diffuse

You will need a photography of some wall, wood, concrete or anything with a good potential to get started. I will use this floor photo from the Psiconic library. Sorry for the black lines on some of the pictures that’s because its a commercial texture.

Here’s our original texture, its not seamless and the light is not even (there’s some shadows remaining). Even subtle changes in lighting across the texture can fool the algorithms used in the softwares like PixPlant.

The reason you need an equalized texture is that you gonna put a light in your 3D application. This light gonna cast shadows and you don’t want shadows with a particular direction in the texture on top of your light shadows. That would not look realistic.

In PixPlant I made a new document of 2048px by 2048px and loaded this texture. I checked equalize lighting and clicked generate. This is the result I got, a 2048px square perfectly seamless diffuse texture !

That’s already a good result. If we load this diffuse in 3Ds Max and map it to a basic sphere here’s the result we can except. Its very flat, there’s not details in the surface, no depth, no specular etc… This could be enough for an object very far from the camera but its not enough for serious photorealistic stuff or architecture visualization.

Adding some details with a normal map

PixPlant can also generate normal maps. They can be used in real time game engine or in any renderer that support them (it’s better than bumpmaps). Here’s the result with some tweaking in PixPlant settings. You can get a real time preview of your material, which is usefull to tweak the settings with a visual feedback. Here’s the output.

And the result of that normal map inside with Mental Ray.

Displacement for large details

You can also generate some displacements maps, that’s great for large details (bricks on a wall for example). Unlike normal maps it actually generate geometry which is more realistic. Depending on what you need you can use only displacement or normal maps or both.

Here’s the result in Mental Ray. The displacement ammount is set to 10 to keep it realistic.

Now if you combine the displacement map and the normal map you get this type of result. As you can see its more detailed than the displacement only.

PixPlant also generate specular maps for your materials. Here’s an example.

Now a render with the diplacement map + normal map + specular map combined in Mental Ray

Now just add the diffuse map and here’s the final image with a displacement of 10.

This example is not perfect because its just the raw output of PixPlant with not a lot of tweaking and there’s no Photoshop work on the maps before or after it. But PixPlant can make wonders if you take time to use it and tweak all the settings. Its also good to input already seamless images in Pixplant because the automatic seamless algorithm doesn’t work for everything, sometimes it’s better to things by hand first

You probably already used Zbrush or seen some screenshots, you can actually apply realtime materials to your models inside Zbrush but it’s all fake. This directX shader for max takes a panorama image in spherical projection and wrap it on your model based on the camera point of view. It’s simple and it works flawlessly.

Let's do this

But before loading the shader itself, let’s grab some spherical panoramas (you can call them matcaps like in Zbrush or litSpheres). One good free source I’ve got come from the Rhinoceros 3D plugin called “Auxpecker” which does the same thing that what we gonna do but in Rhino.

First you need to download Auxpecker for Rhino.

Open the downloaded archive and extract the folder called EnvironmentMap anywhere you want (This folder contain all the spherical panoramas or litSpheres)

Before you jump into 3Ds Max you need to download the directX shader called “Matballz” here and save it anywhere on your computer.

(if you are using 3Ds Max 2012 you need to switch the viewport rendering method from Nitros to DirectX under customize -> viewport -> configure driver to use the Matballz shader correctly)

Now load any object in 3Ds Max and open the material editor.

Pick the DirectX shader in the list
(choose discard old material if prompted to)

Now click on the filepath button and browse to the matballz.fx you previously extracted
(By default 3Ds Max open his own shader folder, put the matball.fx in this folder for quick access)

The matballz.fx shader has very few settings. The only one you should care is the LitSphere map button, click on it and browse to any spherical panorama image file. The file need to have a 1:1 aspect ratio (square)

Here I loaded the “Satined semi-glossy black.png” file from the Auxpecker folder
I used this one for the light bulb body as you can see in the first image of this tutorial

Then you just need to repeat those steps for every different material on your model, here’s another combination with green anodized aluminium and glass.

Using reflection spheres is not only for the look. Its also usefull when modelling, to get more visual feedback of the curvature of your models, its also good for your clients when they are behind you looking at your screen.

You can create or modify the litSpheres with Photoshop or any simillar software, it’s really simple to get impressive results. I hope you enjoyed this little tutorial

(This light bulb model was done in Moment of Inspiration 3D using a Rhinoceros 3D tutorial located here It was done in both software for comparaison purposes. Check the other tutorials on the same website, they are great !)

Le keyframing ne m’a pas pris beaucoups de temps, chaque lettre étant composée de ségments et de pivots j’ai juste animé la rotation des pièces. J’ai placé six caméras dans ma scène et je leur ai assigné à chacune une spline d’animation linéaire en essayant de varier les angles de vue.

Afin de gagner du temps au rendu final, j’ai utilisé la fonction d’export Open GL de Cinema 4D (celle ci va rendre la séquence en temps réel en utilisant la carte graphique de la machine). Avec ces six séquences d’images de prévisualisation converties en vidéos Prores 422 j’ai utilisé Final Cut Pro pour le montage ainsi que pour la synchronisation sonore.

L’avantage de ce workflow basé sur des séquences Open GL c’est que vous savez exactement combien de frames vous aurez besoin de rendre, comme ça vous ne faite pas de rendus inutiles qui prennent souvent plusieurs minutes par images et vous prenez plus soin des autres aspects du projet.

Au niveau du rendu et des matériaux j’ai juste crée deux chromes différents, l’un mat et l’autre brillant. L’environnement est un fichier HDR mappé sur une sphère qui englobe la scène. J’ai été me prendre plusieurs cafés pendants que les images se calculaient en full HD.

Il faut avouer que ce genre d’image n’est pas très flatteuse, un peux trop plat et pas assez contrasté. J’ai donc rendu une passe de zdepth en 32bit en séquence d’images EXR, pour ensuite ajouter de la profondeur de champ dans After Effects en utilisant le plugin Frishluft Lenscare (qui est largement supérieur en qualité que le lensblur inclu dans After Effects par défaut)

J’ai aussi utilisé les plugins Magic Bullet looks et Starglow pour After Effects. Magic Bullet Looks a servi pour la correction colorimétrique et le vignettage sur les bords tandis que Starglow m’a permi d’ajouter des effets de lumière subtiles (light leaks en anglais).

Voila c’est tout. Pour moi ce fut un projet très instructif où j’ai appris beaucoup de choses et notamment comment utiliser une passe de profondeur en 32bit dans After Effects (j’en avais marre du depth-of-field de Cinema 4D qui n’est vraiment pas réaliste).

Video games use them it for all kind of reflections on metallic objects, they also use them for skybox backgrounds (environements) or just special effects on shaders (for example you can fake global illumination using a cubemap) because it’s very fast and the results are impressive.

Recent games even use cubemaps to fake global illumination and diffuse illumination by using a high mipmap level of a cubmap. The high mipmap level is like a blurry version of your sharp cubemap, which contain the light informations but not the details, it’s usually called a “Diffusely Convolved cubemap” (read this article on the polycount wiki for more informations).

Software used in this article

Picturenaut + Filter plugins (read the instructions to install the plugins)
HDRshop v1 (the version two is not a freeware)
ATI cubemapgen (download link at the bottom of the page)

What's a diffusely convolved cubemap ?

Behind this strange word there’s a simple explanation, making a sharp panorama into a diffusely convolved one is like taking a picture in Photoshop and applying some blur. The only difference here, between Photoshop and Picturenaut, is that Picturenaut blur the image and respect the outer edges. The result, when applied to a sphere, has no visible seams !

Sharp reflections or smooth illumination

Here you can see the difference between a sharp panorama and the diffuse version of it. The sharp one is usually used for reflections on metallic objects while the diffuse one is used to mimic the effects of global illumination. You could also have a less diffuse but not quite perfectly sharp one for objects with glossy reflections.

Now, let’s see what’s the result if we don’t use Picturnaut. Here’s a version with some Photoshop gaussian blur. There’s more details, less light from the sun, a lot more brown and orange colors from the buildings. And most importantly (you can’t see it on that image, because it’s flat) there’s a difference between the top and bottom pixels and between the left and right pixels.

How does that look on objects ?

Here’s the previous panorama applied to the Standford rabbit model in a realtime engine. I also added a small ammount of fresnel in the luminance to make the edges a little bit more defined.

As you can see with the diffusely convolved panorama you only get soft lights, in game engine they combine that soft indirect light to dynamic lights (spots, sun, torchlight and so on) to get even more realistic results. You could also combine that version with a slight ammount of the sharp cubemap to fake materials like porcelain.

In the next example I just used the sharp cubemap, you can clearly see the chrome look and feel. Its even more impressive when you move the camera around. Even if the mesh doesn’t reflect itself you get very nice fake realtime reflections. It almost look like a render.

Here’s another example just for fun, the teapot metal is made high mipmap level of the sharp cubemap and the black rubber is made from the same cubemap but diffusely convolved.

Another benefit of using cubemaps

What’s cool about them is that you can have different mipmap levels in the same file (see Mipmap on Wikipedia). Using the embedded mipmap levels you can do both chrome and blurry aluminium reflections using only one cubemap. For sharp reflections like chrome you use the mipmap level 0 (with the higher definition) and for the blurry reflections like aluminium you use a higher mipmap level (with a smaller definition). At each mipmap level the cubemap become blurrier.

Its also good for your graphic card memory, if you look at a chrome object in a video game that’s 10 meters far from the camera you don’t need a 1024 pixel cubemap on it because the object take only 20 pixel on your screen. The object will use a higher resolution cubemap if you get close to it. Think like that, mipmap are like LOD  (see Level of Detail on wikipedia) but for textures.

How do I make cubemaps ?

First of all you need some 360° images, they can be in various projection methods (panorama, spherical, cross). If you don’t know where to get some free images, you can go to the sIBL Archive and download some of their files. In each package you will get a 8bit JPG file and a 32bit EXR file in panorama projection. You can use both.

8bit vs 32bit cubemaps

You can work with HDR images or LDR images (see High dynamic range imaging on wikipedia). LDR images like 8bit JPG are enough for most cases. 32bit cubemaps are used mostly for skyboxes (a good example on the Valve wiki) because they got more dynamic range, thus they give even better results in quality than 8bit images. The only drawback is that they are more expensive for the graphic card. Hopefully you can convert the 32bit files into 8bit files…

Tools to use

There’s not one workflow for cubemap creation, but here’s the tools I use and what purpose they serve. The basic workflow is getting a panorama (or shooting one), converting that panorama to a vertical cross projection and generating the cubemap from that file.

Picturenaut is one of the free tool you gonna love. You can use it to convert the HDR images to LDR ones when you need it. That’s also the tool you will use to generate the Diffusely Convolved version of your sharp panorama.

HDRshop is used to convert your panorama (latitude longitude projection) to a vertical cross one. Here’s an example of a vertical cross projection, it look like an unfolded cube.

Generating the cubemap in ATI Cubemapgen

Once you got a vertical cross projected image you will need to use ATI Cubemapgen. With this tool you can load your vertical cross and filter it to a cubemap. There’s plenty of options, you can generate mipmaps, blur the generated cubemap etc… You can import HDR or LDR vertical cross, DDS cubemaps, or separate faces (one image per face) You can then save DDS cubemap, separate faces or cubemap cross depending on the engine you’re gonna use. I use DDS cubemap since it’s the most common type in real time engines (see DirectDraw Surface on wikipedia).

I usually do a Diffusely Convolved cubemap at 64 pixels per face for lighting. Some 128 pixels per face with different ammounts of blur (for rough materials). And a 1024 or 512 pixels per face for chrome materials with a super small blur of 4 pixels (perfectly sharp reflections look wrong).

Other utilities for Windows

To see the .DDS files in the explorer you need the Nvidia DDS Thumbnail viewer. You can also use the Windows Texture viewer to view your .DDS at full size. There’ also the Nvidia DDS Plugins if you use Photoshop to load and export .DDS files.

Information and other things about this subject

The polycount wiki got one article about cubemaps and another article about the diffusely convolved cubemaps. The Marmoset toolbag is a real time engine to showcase models for game artists which support cubemaps and HDR files as illumination source. It’s also one of the best engine with the Unreal 3 engine for taking screenshots of game assets.

Ressources used in this tutorial

Barcelona rooftops from the Smart IBL Archive on hdrlabs.com (awesome site by the way)
The classic Standford Bunny model from the Standford University, modified version with less polygons by Mr.Bluesummers (good things on his website) smoothed in Cinema 4D.