On Geometrical Properties of Text Token Embeddings for Strong Semantic Binding in Text-to-Image Generation

Open in new window