Paper

TIE: Training-free CLIP Embedding Guidance for Compositional Text-to-Image Alignment · ByteTrail