Refines prompts with parallel or sequential expert agents to boost cultural fidelity in text-to-video generation across mono- and cross-cultural prompts.

github.com/AIM-SCU/CRAFT

arxiv.org/abs/2605.16716