HYPERSKIN
Get UHD heads - On demand!
 |
 |
 |
— Different options exist for creating realistic skin and Hyperskin is one of them. Indeed, a good displacement map is needed in order to render a good looking CG head with a sufficient amount of details such as wrinkles, pores and all imperfections.
The appeal of realism: its specificity
The skin is one of the most, if not the most complicated material to visually reproduce in computer graphics.
The skin is composed of multiple layers, epidermis, dermis, hypodermis but there are also fat, blood veins, hair follicles, muscles, bones etc.. that are inside and under these layers. Moreover, you will find sebum, and many impurities on top: dirt, dead skin... They are only simulated and it is precisely this choice that ends up being the difference between 3D and reality when looking at the final result and references. Over the years, many algorithms were developed such as dipole, directional, random walk (volumetric raytraced SSS)… All of them just take into account the external envelope of the mesh, not the models inside, and this is the backbone of the main issue.
The specular on the other hand works pretty well, microfacet-based BRDFs like Cook-Torrance or GGX help a lot to reproduce the reflectivity and oiliness of the skin. That’s where the micro skin structure is the most visible.
Back to the origin
Character rendering evolved over time thanks to technical and physical improvements such as the increase of screen resolution but also the high number of closeup shots for CGI characters. This in turn resulted in the forced advancement of technology and output in order to keep up with the need to add details.
Back in the day, Artists mostly used procedural noises in order to construct a realistic roughness effect but soon it was not enough to generate a convincing skin.
Today, the standard is now 4K (3840 x 2160 pixels), to put it bluntly: it’s 5 times more pixels than the DVD resolution was ten years ago (720 x 480 pixels).
The bigger the picture/larger the screen, the more detail is needed.
What is the reason behind such a thriving evolution?
Companies are looking for a higher degree of authenticity and realism, they wish to replicate reality as much as possible. Nowadays, this trend has become a must-have 3D standard as the need to produce better visual aspects for the skin is becoming greater and greater.
Uncanny valley: the insurmountable hurdle?
When an object’s resemblance to a human is nearly faultless, the human mind perceives the tiniest details that can betray the fabricated illusion. This emotional response is hard to counter as the theory suggests that humanoid subjects, which are imperfect replicas of human beings, prompt the observer to develop feelings of eeriness.
Skin’s complexity
Companies seek to reproduce the skin appearance in the most faithful way. But to do so, it is important to understand how difficult skin is to render correctly. Indeed, as human beings we are acutely sensitive to the appearance of skin: it is the sum of many subtle visual characteristics that makes the difference between a convincing skin and a plastic doll.
That’s why a genuine model of skin must include a sheer quantity of features such as wrinkles, pores, scars, hair follicles, to name a few of them. It is a complicated process but Hyperskin was designed to solve a part of this barrier. The displacement is essential to get the skin to look true to life. Without that, you usually get a specular close to plastic or we get the feeling that something is missing.
How to bridge this gap ?
Workflows had to evolve throughout time, especially when taking into account the breakthrough of the 3D scan technology but also the emergence of resources such as TexturingXYZ maps. It is now a standard that enables countless companies of the industry to replicate faces’ 3D models and textures as similar as possible. Most high-end photometric scanning systems, such as a Light stage captures textures at 100 μm of detail, which get most of the “meso” structure.
However, even with the help of scans, you have to clean and add details for the skin to appear lifelike.
It’s the reason why today 2 workflow standards coexist:
— Sculpt & textures projection
Both complement one another in regards to their pros and cons: they are efficient for adding details but sculpting software has polys limitations (except when using HD geometry mode in Zbrush) and is time-consuming whereas the texture projection workflow could take less time but need good quality images to project. Both can introduce some consistency issues which depend on the Artist’s skills and resources available.
What Hyperskin can do for you?
Our dedicated technology Hyperskin places itself as a perfect solution against one key barrier preventing the skin’s appearance to look realistic. With a single displacement map in 16k, you could get details down to 10 μm.
— With UDIMs, we could get with ease 1 μm skin details.
Renders produced with a highly detailed displacement map will be far more accurate since the model’s reaction to the light will be better. Thanks to the output produced by Hyperskin, the new displacement map(s) will help the SSS and specular to behave to a higher standard, and allow them to render images that better represent reality.
To put it in a nutshell, realistic skin is still a complex ideal laborious to obtain when thinking about the difficulty introduced by the uncanny valley phenomena. So when a 3D character appears to look a little fake, this perception comes from the difficulty to overcome several roadblocks along the way at different levels: visual, animation, and so on.
What matters in CGI is how the representation of matter relies on various components and to optimize the process to get the best results possible.
