Our work MaxFusion was accepted at ECCV 2024. MaxFusion enables training-free multimodal spatial conditioning in text-to-image diffusion models.