Natural Redesign for an AI Generative Tool

Challenges and Solutions

INTRODUCTION

In this project, I took on the challenge of redesigning the old version of the Stable Diffusion WebUI during my time at ByteDance. This AI image generation software was highly regarded by creative individuals for its open-source flexibility and customization features. The goal was to quickly create an immediate solution that could be showcased to senior leaders as a demo—proving not just an improved Stable Diffusion but also our capability to deliver a better user experience for similar AI-generated content (AIGC) tools.

Let's Connect to See More Details!

Because of a confidential issue, I am not able to present deeply about what I’ve done. If you are interested in my journey, please contact me to see how I tackled the challenge of redesigning the interface, solving immediate problems, and delivering a seamless user experience under tight deadlines.

Send me an email!

Step 1: Competitor Analysis to Understand the Market

By looking at popular AIGC image generation products, I noticed that while they all offer text-to-image generation, they differ a lot in accuracy and control. This includes factors like available plugins, model resources, refining options, and user-friendliness. Even though Stable Diffusion WebUI has great customization features, many users still struggle to use it effectively.

SO WHAT WAS THE PROBLEM?

High flexibility leads to a steep learning curve and increased operational complexities.

The old version had many powerful features, but users struggled to find them. Even when they did, it was still tough to use effectively, especially for those without industry knowledge.

Step 2: Dive Deeper to Understand Target Users

By gathering feedback on popular AIGC products from well-known forums, I create user personas that highlight the different needs users face in various work scenarios.

Step 3: Identify and Refine Pain Points Through User Journey Mapping

To better understand users' needs, I conducted interviews with 9 colleagues—both experienced users and novice users. I summarized the identified pain points and optimization opportunities in the table below.