AI agent improves control over positive and negative space in images
Form and Void: Entangled Composition through an Autonomous AI Agent
Computer Vision and Pattern RecognitionMultiagent Systems
Summary
Creating images that balance objects and the spaces around them is hard because both parts need to fit together nicely. The authors designed an AI agent called FaV-A that makes this easier by first drawing the main shape, then looking at it to find meaningful empty spaces, and finally giving instructions to complete the image. Their tests show this step-by-step approach works better than trying to do it all at once with existing AI models. This helps make pictures that look better and match their meanings more closely.
What this means in practice
- •For graphic designers: Create images with better balanced shapes and background spaces by using staged AI instructions for clearer visual storytelling.
- •For advertising teams: Generate visually coherent ad graphics where positive and negative spaces are controlled to enhance message clarity and impact.$Commercial implications: Enables AI-generated ad visuals with refined composition, improving creative workflows and resulting in more effective marketing materials.
Authors
Shiwen Wang, Jian Yang, Xu Wang, Xincan Wang, Weiming Dong
Abstract
Positive and negative space is a fundamental principle in visual composition, supporting visually coherent forms and layered semantic relationships. Generating such compositions is challenging because it requires coordinated control over two semantic concepts that share a common boundary. Although recent text-to-image models and multimodal large language models (MLLMs) have achieved strong performance in image generation and visual understanding, positive-negative space generation remains difficult, particularly under direct single-pass prompting. In this work, we present the \textbf{F}orm \textbf{a}nd \textbf{V}oid \textbf{A}gent (\textbf{FaV-A}), a multimodal agent designed for staged positive-negative space generation. FaV-A follows a progressive workflow: it first generates a base object, then analyzes its shape and spatial structure to identify candidate negative-space semantics, and finally produces compositional instructions for the final image generation stage. Experimental results and ablation analyses suggest that FaV-A provides a more effective framework than direct zero-shot MLLM baselines for producing visually coherent and semantically aligned positive-negative space compositions.