Research map / Image generation / Text-to-image diffusion
Versatile Diffusion: Text, Images and Variations All in One Diffusion Model
Xingqian Xu, Zhangyang Wang, Gong Zhang, Kai Wang, Humphrey Shi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| mix image and text contexts at attention level instead of layer level (ablation) | VD dual-context blender · dual-context blending (qualitative) | visual quality | Improved |
| use VD's layer-level multi-flow mixing instead of model-level ensembling of SDv1.4 and SD-variation (ablation) | dual-context blender · dual-context blending (qualitative) | visual quality | Improved |
| train with text-to-image as beginning task (VD-Alt) instead of image-variation (ablation) | VD · text-to-image and image-variation (qualitative) | visual quality | No clear effect |
| use two-flow or four-flow instead of single-flow model for disentanglement (ablation) | VD · style/semantic disentanglement (qualitative) | disentanglement quality | Improved |
| remove the global CLIP image vector from the context embedding (ablation) | VD image variation · image-variation (qualitative) | image structure | Hurt |
| use all-zero unconditional embedding instead of CLIP embedding of empty image (ablation) | VD image variation · image-variation (qualitative) | robustness on photorealistic inputs | Mixed |
| use normalized and projected CLIP embeddings as context | VD context encoder · training | convergence and quality | Improved |
Closest papers by meaning
- Dual Diffusion for Unified Image Generation and Understanding (CVPR 2025 · similarity 0.90)
- Diffuse Everything: Multimodal Diffusion Models on Arbitrary State Spaces (ICML 2025 · similarity 0.90)
- Omni-Diffusion: Unified Multimodal Understanding and Generation with Masked Discrete Diffusion (ICML 2026 · similarity 0.90)
- CONFORM: Contrast is All You Need for High-Fidelity Text-to-Image Diffusion Models (CVPR 2024 · similarity 0.90)
- Diffusion PID: Interpreting Diffusion via Partial Information Decomposition (NeurIPS 2024 · similarity 0.90)
- I Think, Therefore I Diffuse: Enabling Multimodal In-Context Reasoning in Diffusion Models (ICML 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
