Automated asset design for brand development & Consistency

Sawa identifies familiar graphic design practices and develops an interactive system based on these practices. The outcome consists of three contiguous steps: Setting Constraints, examining Variations and Iterating through compositions.

Overview | Credits



A sense of progression
Design tools like Adobe Suite have made graphic design ubiquitous and accessible to a much wider audience. They represent initial technological advancements that automate the process of designing.

Templates as a first step
Templates have been offered as a solution to absolve the designer of these issues. They allow those not as skilled in design to create a professional looking product, because a designer created the template, and all the next person did was fill it in with their own photos or text.

The result of templates is a lack of identity. Most users find this problematic, as most brands do not want to look like other brands. This issue has evidently been acknowledged with the advent of suggestive interfaces, even within tools like Keynote. 

Moving beyond templates
Sawa proposes a system by which we can further assist anyone in developing uniquely tailored design outputs. This proposal relies on the notion of Suggestions as a tool to improve the design process. Suggestions can exist in multiple forms: Variations, a set of distinct, alternative solutions to a given problem at a point in time. Iterations, versions of the same solution at different points of revision. 


To help anyone specify their intent, I introduce additional constraints in the model that first define the dimensions of the canvas. These are initially explicitly specified, but later inferred from user behavior over time. 

The system begins with a set of design compositions that structurally differ in both style and layout options. Be different, anytime. Simply pinching to zoom out showcases variations. The system assumes variations exist on a randomized canvas that implies variety; a space in which users can explore visibly distinct compositions that aren’t systematized.

Iterations represent small incremental changes for the same composition. The system acclimates to user inputs, generating new iterations as changes are made to compositions. The system perceives iterations as either corrective, when compositions are inaccessible, or entirely suggestive, when the system simply follows common practices.

The purpose of this work was to help provide an informed direction to product managers and executives. Majority of the foundational theories presented in this project resulted in the launch of The Creative Assistant, Mailchimp's first AI creative tool.