DreamPoster: A Unified Framework for Image-Conditioned Generative Poster Design
Xiwei Hu, Haokun Chen, Zhongqi Qi, Hui Zhang, Dexiang Hong, Jie Shao, Xinglong Wu
2025-07-15
Summary
This paper talks about DreamPoster, a new system that uses text and images to create high-quality posters with flexible sizes and designs using a tool called Seedream3.0.
What's the problem?
The problem is that making posters with AI can be difficult because existing methods can’t generate high-quality images that fit different sizes and layouts well, limiting creativity and usefulness.
What's the solution?
DreamPoster solves this by using a framework that combines text instructions and images to guide the generation process, allowing it to create posters that look great and can be adjusted to different layouts and resolutions easily.
Why it matters?
This matters because it makes it easier and faster for people to design custom posters with AI that meet their specific needs, helping artists, marketers, and creators produce better visual content.
Abstract
DreamPoster, a Text-to-Image framework using Seedream3.0, generates high-quality posters with flexible resolution and layout, outperforming existing methods.