Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation

Open in new window