Scaling AI the Smart Way: Memory Footprint and Runtime Efficiency | Unreal Fest Chicago 2026

As games increasingly adopt real-time inference, memory footprint and runtime efficiency have become just as critical as raw model quality.

This talk recorded at Unreal Fest Chicago 2026 explores how Databiomes has worked to optimize AI models for practical, in‑game runtime use—balancing fidelity, performance, energy efficiency, and platform constraints, through close collaboration with AMD.

Databiomes and AMD walk through key lessons learned from adapting offline‑trained models for real‑time execution in Unreal Engine, including strategies for reducing memory pressure, improving latency, and scaling across a wide range of hardware CPU, GPU, and NPU.

The session also highlights how AMD technologies and plugins can help developers achieve additional runtime savings, enabling AI features that fit within real‑world performance and memory budgets. Watch to come away with a clearer understanding of what it takes to ship AI in games today.

Find out more about using Unreal Engine to develop games here: unrealengine.com/uses/games

#Art, #3DAssets, #Tripo, #NanoBanana, #ChatGPT, #UnrealEngine, #UnrealFestChicago2026