The task consisted of taking the Houdini project we originally built to generate 11,000 PFPs:
https://opensea.io/collection/parallel-avatars
and transforming them into fully playable characters in Unreal Engine for iOS and Android.
I was responsible for developing the entire character pipeline inside Unreal Engine, including character assembly, skinning, shaders, physics, and performance optimization. This involved generating dedicated low-poly character variants and implementing systems to reduce draw calls and improve runtime performance on mobile devices.
I also created the Blueprints, CSV-driven data pipelines, and all the underlying gameplay and customization logic required to make the system work efficiently on low-end Android devices.
In addition, I implemented a complete facial animation system for every character, including facial morph targets, facial animation generation, and the runtime emotion-response framework. The characters could converse with players through an LLM-driven dialogue system, and their facial expressions would dynamically react to the emotional state inferred from the conversation, creating a responsive and expressive character experience.
I worked on the facial animation system from the technical side, including Blueprints, implementing the morph target system, generating animation tracks, and balancing all the systems to improve the lip-sync experience. Additionally, the system includes s
Here's the character customization system. I designed the entire pipeline from Houdini to Unreal Engine, overcoming the technical limitations and finding the most efficient solutions to ensure compatibility with low-end devices.
Facial animation tests.
I had to implement a set of special characters that were built very differently from the rest of the Houdini-generated PFPs, while still using the same system that powered all 11,000 PFPs for optimization and performance reasons.