PROPOSAL
Company: Mythica
About:
Mythica wanted to use AI/ML loops to allow Environmental Artists and Level designers to upload 2D concept art and generate 3D vegetation assets for faster development in video game cycles
​
The Problem: Level Design cycles in game development take too long and are expensive. Can AI/ML make the process faster while keeping artists at the forefront of level design?
​
Role:
Product Design Lead
​
Duration:
3 months
RESEARCH
To define the architecture of the plugin, we asked our Level Design users to define what are some of the most important questions that come to mind about their level building process. Those were shortlisted to top 3:
​
1. What is their current process to blockout a 3D level in a game? How do they go from from napkin sketch concept to greyboxing a level to a low fidelity prototype?
2. What are their pain points while modeling in Houdini to Unreal Engine?
3. How close is their concept art to the final level design? Could it be closer?
From the feedback, the CEO, key stakeholders, and I ideated in Miro how to address each pain point. The result was a Houdini pipeline inside Unreal engine where a user could upload concept art to create greyboxed levels using image segementation and RLHF loops.
PROTOTYPING
Using Figma and Miro, I suggested a flow where level designers could upload concept art and produce a level quickly. The plugin would pull in models from Houdini for ready use in Unreal Engine, generating the models based off of the concept art, keeping the designs custom to the game without stealing artwork from artists.
​
Project PW: Scully24 (case sensitive)
GAME DEVELOPERS CONFERENCE
Deliverables.
Here is the trailer of the user flow I designed, animated, and produced for GDC in San Francisco.
RESULT
Key learnings
We saw success in early tests among Game Developer and Level Design users and moved onto designing an additional website for designers to upload their assets into a cloud-based, open source experience.
​
Major takeaways:
​
Designers were less hesitant to use AI/ML models once their work stayed custom and in-house
People wanted faster ways to do their job, letting the machine do the grunt work while they focused on creativity
Game devs wanted an assistant - not a replacement