Sanguine

Sanguine is an interactive Art Project / Video Game, that I worked on over two Semesters at the HdM Stuttgart (Link to HdM Project Site).

Our goal with this project was a game about belonging, familial bonds, transformation and freedom, where the player ultimately had to choose which way they would want to take, deciding for a helpless and deformed being and see where their choice lead them.

Sanguine Trailer:

The project was done over two semester, for which the first was used for building a concept and prototype. For the second semester the focus shifted a bit and the team grew considerably.

First Semester

In the first Semester we built a prototype of the game, which can be seen below:

My part in this was:

• Modelling and Retopology
• Lighting Concept
• Coding / set up Unreal Blueprints mechanics
• Machine Learning a model to process user text input to text intent
• User Interface

Modelling / Retopology

I did the complex Retopology of the main character’s head, which was a fun challenge to make this complex form fit the structure of quad faces…

Coding & MAchine Learning

Originally we wanted the player to be able to ask anything via text input, which we wanted to process with a program, breaking it down into words it could detect from a pre-written list, which an AI model I built would try to match with the question it thought was most likely being asked. For the question the AI set as the most probable, we could then draw from a selection of pre written answers, which we wanted to be cryptic anyway, so even if it couldn’t match the text input 100% perfect all the time, the cryptic answer should’ve still been quite immersive and would’ve enabled us to user voice lines for the answers.

I trained the model with Python and because we wanted it to work within Unreal, which is based in C++, I wrote a C++ Program which could access and run the Keras model. Sadly the work of implementing it was too big, so we changed direction to a dialogue tree in the second semester. In the demo above we put in some very specific questions and answers to showcase the functionality we were aiming for.

In the following showcase you can see how the C++ program breaks down the input, cross-references it with the list of words and feeds it to the Model which then predicts the most likely question being asked

Second Semester

In the second semester I worked on the masterlight for the lower path, as well as the shot lighting for the ending cinematic of the lower path.

It was conceptualised as otherworldy, surreal, which I also implemented within the lighting. Later there was the creative decision by the director to go towards a greener tint to not have the setting feel too comfortable.

Here the concept art for this setting (not by me):

This was an early version of the Lighting:

and here is the final Masterlight for the lower setting:

Below is the ending cinematic of the lower path, for which I did all lighting: