Art of the Possible: Experimenting with AI in Art (Presented by Keywords Studios)

This session demystifies how AI can reliably produce high‑fidelity art, animation and video assets for live game production. It begins with "brand‑trained" pipelines that ingest existing assets and learn a studio's visual language; these pipelines handle data curation, model training, evaluation, inference and operations. Examples include stylized avatar creation, novel character and IP generation and synthetic dataset production. The talk then explores a Swiss‑army‑knife toolkit: LLM‑driven tools, multi‑view 3D reconstruction, multi‑modal character generation using segment prompts and image‑likeness matching, and gameplay‑footage analysis that automatically selects segments, composes trailers and adds AI‑generated audio. It covers personalized video flows, metahuman localization and motion capture via text prompts. Throughout, the emphasis is on responsible adoption, knowledge‑sharing and staying ahead of rapid technological change.