Opens with a description of the engine. Apparently the demo is based on the concept that while they aren't played by actors in the traditional sense, video game characters still have costume changes, so they made a scene based on that premise. The writer then comments on the high detail in reflections, eyebrow hairs, and general lighting.
There is a specific tech in the demo dubbed Luminous Hair that enables highly detailed hair for characters. They also have carried over for this demo the ability to change clothing and hairstyles on the fly, something incorporated into FFXV. The current challenges they've surmounted with Luminous include environments and characters' hair and clothes. Next on their list is the ancillary details such as skin complexion. Due to this workload, the team are working with several outside collaborators, including a New Zealand university who works with them on "CG" and "VR" elements, and Kyoto University for help with the technical maths.
They then include several examples of the current engine's capacities. In regards to skin colour, they recognise that it has a wide range, and want their tech to be able to reflect both general colour and surface imperfections. They also sought a way to "mechanise" the creation of both super-high def and lower-def models, as the problem with handmade models is that the polygons can't be reduced without difficulty. Their geometry processing can now cut unnecessary polygons while retaining a model's traits and shape.
Next, he talks about the demo's production. There were several intertwined production elements, very closely linked. The woman's head was a model that could undergo slight adjustments to expressiveness based on markers. The model has a CGI base, then a real-time version with the bones necessary for forty different expressions before all the textures are laid on. The team then adjusted how expressive the character had to be and how the clothing moved. The current system is a substantial expansion of the character creator from FFXV Comrades, in that the basic tech now uses multiple reference points to create a character rather than a binary system of "source" and "target shape". On Luminous Hair again, they've basically taken their Maya hair tech and let it off the leach in terms of adjustable details, plus adaptability to a number of different body types without compromising the style.
In closing, the team say the display build uses these things as manual system processes. They will now automate them, with the Luminous Hair adaptive tech also applying to clothes. This means they can considerably cut the development time spent on graphical assets.
(There, that's the gist of it.)
