Military Utilities

This small project was a collaboration between myself as the artist and my partner, Yiğit Sakin, as the AI developer. We’re trying to come up with ways on how to produce concept art using AI solutions. We’re working on our own stable diffusion model that’ll produce matte painting concept art pieces based on blockout scenes/models and respective depth maps that I provide.

Our goal is to train our AI model so that it can texture, conceptualize, compose concept art with basic meshes. We plan to use the AI to gives us perspective on how we can iterate our projects and improve our visual outputs.

This project was our first initial test. I’ve provided an extremely basic scene, and a depth map for that scene (Both the basic render and the depth map are provided at the end). Our AI model produced several concept art pieces of which I’ve chosen 2 as my primary reference for lighting and composition (Both concept art pieces are provided.

My goal was to use only the concepts provided by our AI and nothing more for lighting and composition. I was trying to gauge whether the lighting was accurate enough for me to base my scene, or If I’d need additional references.

The results were satisfactory. I only used the AI concepts and didn’t need any other reference to base my lighting on.
We believe the initial steps were positive and we’ll continue to work on our AI concept art generator.

PS: All the 3D assets (Model and texture) were created completely by myself, not the AI.

TOP LEFT: Render given to AI, TOP RIGHT: Depth map given to AI, BOT: Concepts produced by the AI (Light and Dark version)

TOP LEFT: Render given to AI, TOP RIGHT: Depth map given to AI, BOT: Concepts produced by the AI (Light and Dark version)