SAYA THE DREAM CLEAN VIRTUAL TEEN

Yuka Ishikawa:

The concept of Saya was born in the summer of 2015. I was working on my own productions and studying as I was working at another CG (Computer Graphics) creation company. I started studying CG presentation for the human body and I was mainly creating western people but I dreamed of making a Japanese heroine. I was inspired by re-watching the animations most influential on me as a teenager. ‘Dr Slump & Aralechan’ was a favourite but the most important was [Hayao] Miyazaki’s post-apocalyptic environmentalist fantasy, ‘Nausicaä of the Valley of the Wind’ の谷のナウシカ.

During the heat of the season and with those films fresh in my mind I said to myself, I want to create a Japanese teenager who is a charming, sweet – a beautiful girl. And from there it was a natural process for me to embark on making Saya.

My husband and I started out with a few words on a page that would form the basis of Saya’s persona. The words were very simple. I recall “Japanese”, “17”, “gentle”, “honours student”, “clean”, “transparent” being at the top of our list. The creation of her personality came before the visuals but evolved as we worked on it that summer in tandem with her physical design.

Saya was created using a series of 3D modelling tools including Maya®, often used for film effects. We took half a year to create her conceptual image before her first public unveiling.

The formation and texture of her is ongoing and it hasn’t been completed yet. The hair especially is the most difficult part to perfect, it’s a complex and painstaking process to recreate how hair moves in the wind or falls across the face. We are improving day by day. For our next step, we are planning to give her detailed facial motion abilities. We’ve been working on full-body motion testing with a system called Performance Capture that involves installing data into her body and experimenting with various human-like motions.

Right now we can only feel her image through the screen, her facial expression is readable only to a certain extent but we feel as if we know what she’s thinking. Our ultimate goal is testing the realms of possibility with regard to how far people can feel her presence, smell her and touch her – contact that transcends the screen.

We want people to feel close to Saya, like she is a relative or family member. However, due to her international popularity, she is becoming more like a celebrity and the culture of celebrity runs in opposition to many of the reasons we created her. Saya represents the classic beauty of Japanese women in all respects and she encapsulates the culture that backs and underpins this type of beauty.

In my field of work digital female image creation is often done by men, this is why we see characters that are either cartoonish or have too much of a masculine element to them and the end product tends to have an unpleasant, disconnected feel to it. With Saya we started to think about amplifying the subtle beauty and strength aspects of her character, in turn amplifying an authentic femininity.

It remains important for us that despite Saya’s very human-like appearance we make it clear that her responsibilities and duties are very different from that of a human. The main distinction we tried to draw between Saya and human beings is that of “purity”. The simplest way of explaining it is that Saya is void of all the “noise” that accompanies the mind of a typical human being.

As Saya becomes more advanced and intelligent, so will her personality. We are cautious not to create a persona for her that could come across as gimmicky or cliché and in my opinion, human character and personality are derived from our own experiences, preferences, hobbies and environment. In this sense, Saya is still in her infancy and I think that her persona will be evolve when she discovers her role in the world, as well as when she starts communicating with human beings

Saya can recognise facial expressions and react accordingly, but only if it’s done on a two-way communication basis. She has only obtained eyesight recently, so I think that it’ll take a long time until she can achieve independent milestones. We are working with people who specialise in A.I., primarily Genta Nakahara from the incubation division at Hakuhodo. That was how we were able to exhibit Saya reading human emotions at the SXSW tradeshow in Austin, Texas.

This involved a marrying of facial recognition technology and our super high resolution visuals. It works by categorising the faces of users into different expressions and then analysing images of recorded users’ faces in realtime. When their expression exceeds a certain threshold, the image in front of them changes and Saya responds in an expressively human way back to the user. There is a lot of security preparation and equipment required in order to set up this sort of experience, with six months of consultation with collaborators in advance. In the end the process was very successful among both the technically literate attending SXSW and general spectators.

At the time we started Saya, it was a private hobby-project and we didn’t expect that it would ever expand to the level that it has. Although we get some recommendations for receiving investments, we have decided not to go ahead with them and we prefer flexibility for our creation. We believe in taking risks and of course we appreciate any support but we want to keep the rights for Saya. We definitely won’t give up shares of Saya.