Empowering Backbone Models for Visual Text Generation with Input Granularity Control and Glyph-Aware Training
Generating accurate and aesthetically appealing visual texts in text-to-image generation models presents a significant challenge. While diffusion-based models have achieved...