Harmonized Banner Creation from Multimodal Design Assets

ACM CHI Conference on Human Factors in Computing Systems (Late Breaking Work)

Published April 30, 2022

Praneetha Vaddamanu, Vinay Aggarwal, Bhanu Prakash Reddy Guda, Balaji Vasan Srinivasan, Niyati Chhaya

Designing aesthetically pleasing single-page graphic designs (”banners”) that appeal to the target recipients is non-trivial and requires considerable human effort. Designers often start by collecting assets like background images, inspirational banners, and relevant taglines/textual phrases, and iterate over variants of the designs by testing different combinations of these assets for their current task. To expand creative processes beyond professionals, it is crucial to accelerate this tedious process of creating design variants. To this end, we propose an algorithm that takes as input multiple image and text elements to generate several visually pleasing banner designs harmonized for text colors, font styles, and the placement of these multimodal assets. The generated layout accounts for background image saliency and uses this to select typographic and color variants yielding the harmonized outputs. We demonstrate the effectiveness of the proposed algorithm in generating aesthetic banners with the help of a crowdsourced experiment.

Learn More