Through our work process, we have managed to reap the full benefits of the pre-calculated time by working with real time. Working in real time saves a lot of time during project production. This is not the only advantage as the assets composing the different scenes can be reused for a virtual visit, for example. Another 3D studio would have to recompose the scene in full for each destination platform.
Working in real time allows you to see your scene in a version which is very close to the final result.
The computation time for rendering an image is only a few minutes.
The result obtained is very close to the version during the stage composition. There are very few changes to make and therefore a lot of time is saved.
· You immediately see what your scene will look like when you compose it.
· Quick computation time.
· No need for a lot of calculation servers.
Given the long calculation time, we lose more time upstream to prepare the scene well and avoid making too many changes afterwards.
Depending on the complexity of the scene, computation time can take up to a full day for a single image.
Once you have seen what the result looks like, if you are not satisfied, you have to go back and make the corrections and wait again for the rendering time until you are satisfied with the result.
· You need to calculate your renderings to know what your scene will look like.
· Much higher computation time.
· You need a lot more calculation servers.
The use of animation physics makes character movements much more realistic. It is the same for the behavior of objects, during a fall for example.
Physics can be used for many other things such as hair, fur, vehicles, anything that can be animated and subject to the laws of physics of our world.
In this example, the lantern is set to move properly, according to the laws of physics, if you push on it. The rocking of the object will take into account its weight and its center of gravity.
Each hair or hair is placed independently of the others, and therefore reacts much more naturally.
Their movements not only follow those of the head or body but also change with the wind and any other external factor that could affect their immediate environment.
A vehicle will look much more realistic if we apply physics to the suspensions, it will have a more natural movement during a jump, a bump pass or a turn.
"PBR" stands for Physically Based Rendering and describes a collection of 3D scenes rendering techniques imitating physical models that describe lighting conditions in real world.
For a physical material to appear realistic, it is imperative to recover physical coefficients in an exact way, something that very few studios on the market can do at present. Thanks to our working processes we recover all these maps, in a physical way and with exactitude, which gives a very realistic aspect to our materials.
· Ambient Occlusion : darkens naturally difficult areas of light.
· Albedo : unlighted material colors.
· Gloss : low roughness of the material.
· Height : material relief.
· Metallic : metallic appearance of the material, anisotropic coefficient.
· Normal : material relief.
· Opacity : material transparency.
· Roughness : intensity of the asperities of the material.
· Specular Level : way the light bounces and is absorbed by a material.
· Subsurface : way the light passes through the material.
Currently most 3D studios in the world, make a rough estimate of these properties and create them manually. Especially for Specular values, Roughness and Freznel effects. Thanks to our captation techniques and our expertise in photogrammetry, we are able to recover all the exact coefficients of each of the maps of all materials, and thus create textures identical to reality.
Our colors are very finely calibrated and transcribe accuracy to reality.
On the left side of the picture, the original woolen sweater; and on the right side, our texture.
When creating a texture, it is very important to eliminate any light reflections that were inevitably made when scanning the original subject.
The texture you create needs only the reflections of the light that are placed in the 3D scene. In the opposite case, where there would be residual reflections during the scan, the texture will look completely wrong.
On the left side of the photo is the "Map Albedo" which still contains light reflections. On the right, the same map after our treatment process.
We have implemented artificial intelligence processes that automate work steps and provide much better performance.
The process of scanning an object to recover its volume and texture is called "photogrammetry".
The artificial intelligence that we developed allows us, for example, to "delight" the textures and to replace it in the right place automatically on the 3D volume of the object.
In the case of a more generic texture like that of a stone, we can also use the texture of a stone to replace it automatically on another while respecting the volume and relief of it.
The object will therefore react to the light to which it is correctly exposed.
Artificial intelligence is also used in animation.
If we take the example of a character who is walking and accelerating to start a race, artificial intelligence predicts how the animation should be made to make the transition natural.