The field of text-to-image synthesis and editing is rapidly advancing, with a focus on improving the control and fidelity of generated images. Recent developments have introduced new methods for incorporating negative prompt guidance, style-specific content creation, and anomaly generation. These innovations have led to significant improvements in image quality and adherence to textual prompts. Notably, the integration of large language models and diffusion transformers has enhanced the understanding and execution of complex instructions. Furthermore, advances in autoregressive modeling and flow matching have enabled more precise and efficient image editing. Overall, the field is moving towards more sophisticated and controllable image synthesis and editing capabilities. Noteworthy papers include VSF, which introduces a simple and efficient method for negative prompt guidance, and DeCoT, which leverages large language models to enhance text-to-image generation. Additionally, papers like SAGA and CurveFlow have made significant contributions to improving the fidelity and control of generated images.
Advances in Text-to-Image Synthesis and Editing
Sources
VSF: Simple, Efficient, and Effective Negative Guidance in Few-Step Image Generation Models By \underline{V}alue \underline{S}ign \underline{F}lip
SPG: Style-Prompting Guidance for Style-Specific Content Creation
Training-Free Anomaly Generation via Dual-Attention Enhancement in Diffusion Model
Demystifying Foreground-Background Memorization in Diffusion Models
DeCoT: Decomposing Complex Instructions for Enhanced Text-to-Image Generation with Large Language Models
Single-Reference Text-to-Image Manipulation with Dual Contrastive Denoising Score
Leveraging Diffusion Models for Stylization using Multiple Style Images
7Bench: a Comprehensive Benchmark for Layout-guided Text-to-image Models
Text2Weight: Bridging Natural Language and Neural Network Weight Spaces
DegDiT: Controllable Audio Generation with Dynamic Event Graph Guided Diffusion Transformer
SAGA: Learning Signal-Aligned Distributions for Improved Text-to-Image Generation
CTA-Flux: Integrating Chinese Cultural Semantics into High-Quality English Text-to-Image Communities
MUSE: Multi-Subject Unified Synthesis via Explicit Layout Semantic Expansion
SATURN: Autoregressive Image Generation Guided by Scene Graphs
TransLight: Image-Guided Customized Lighting Control with Generative Decoupling
Hybrelighter: Combining Deep Anisotropic Diffusion and Scene Reconstruction for On-device Real-time Relighting in Mixed Reality
CurveFlow: Curvature-Guided Flow Matching for Image Generation
Visual Autoregressive Modeling for Instruction-Guided Image Editing