HYPERSKIN
Get UHD heads - On demand!
![]() |
![]() |
![]() |
In our beloved industry, the appearance of the skin is considered to be a very important topic. Models with skin-like appearance are commonly used to render digital human characters in productions.
For these “digi doubles” to appear realistic and be fully integrated into a shot, it is essential that their skin appearance precisely captures all the subtleties of actual human skin under different point of view and lighting conditions.
Uprez, enhance, upscale… you might be familiar with all these terms, most of them are used for 2D images nowadays although they can also be used in a 3D context, as soon as it improves the overall quality of an asset.
Render based on Digital Emily 2.0 - Copyright ICT| Wikihuman project |
The need
We had many requests from VFX and Game companies to “uprez” 3d head models that already had a lot of details, coming from ICT’s lightstage data or equivalent, that produces a super high-resolution 3d model based on both photogrammetry and photometric acquisition.
You might wonder why an HD model would need an uprez;
- Such technologies can capture the skin detail to ~10µ precision, called mesostructure or meso scale. Mostly details that are visible to the naked eyes, at a certain distance from the subject.
- The photometric capture could produce different output quality, if the subject and/or camera moves between takes, even just a little (1px shifting is enough), then the details won’t be well captured as the images won’t be 100% aligned for the normal computation.
- Many other factors could compromise the quality such as vellus hairs, very light skin tone, or oily skin to name a few.
Render based on Digital Emily 2.0 - Copyright ICT| Wikihuman project |
Pores and wrinkles are observed on the surface of skin layers. They make the structure of skin rough and consequently contribute to the diffusion of incident light at the surface of skin.

Render based on Digital Emily 2.0 - Copyright ICT| Wikihuman project |
We can also talk about gear and optics, even with high-end and very costly cameras, you would need very good optics as well to capture all the micro details. There is still a limitation based on the sharpness and resolution that could capture the scan rig.
With the mesostructure captured, there is still a huge gap and many things missing between this level of detail and the micro details that could produce 1/1 digi double.
A perfect digital twin must have all the wrinkles and pores well defined, all over the face.

Render based on Digital Emily 2.0 - Copyright ICT| Wikihuman project |
Beyond the High Definition standard
— You have probably guessed, it’s now time to show the potential of Hyperskin to uprez a model that has a lot of details. For this use case, we wanted to share the results based on the very well-known and great 3D asset provided by ICT and the Digital human leagues, Digital Emily 2. This model shows very well the level of details that are captured by the pioneer in this field. The goal of this study is to present the work we did using Hyperskin to enhance the model.
The data
With Digital Emily 2 project, we also have photography references, Cross Polarized Point Light, Parallel Polarized Point Light, and Specular Only Point Light data. This is the ideal case as we can see the result using the provided displacement and the newly generated one with the help of Hyperskin.

Copyright ICT| Wikihuman project |
The moment of truth
Preserving everything while adding consistent details, working with and not against the existing data, that’s what we do with Hyperskin.


Render based Digital Emily 2.0 - Copyright ICT| Wikihuman project |
Important notes: In the references provided by ICT, the highlights are clipped and contain some DOF. We also had to guess the light position, so the comparison isn't perfect.
Without any doubts, you have seen the difference between the before and after hyperskin process! Unlike others, our uprez process doesn’t rely on AI, this shows the flexibility of our technology and how far we could go with it.
We push the standards of quality of digi-double assets.
References:
Digital Human Leagues | Wikihuman project
USC ICT | Paul Debevec