Timeline

September - November 2024 (12 weeks)

My Role

UI/UX Design for Community Work

Problem

AI image generation platforms can overwhelm beginners with complex prompts and frustrate experienced users with fragmented workflows. The lack of visual previews and interactive tools creates uncertainty, while limited task management features hinder progress tracking, reducing efficiency and user satisfaction.

Business Model

Solution

"Smart" Interface: Predictive and Adaptive Generation Based on User Actions

Boosting User Learning and Platform Familiarity by 3x with Visual Tools and Keyword Assistance

Introduction

I hosted three user interviews using Figjam and sticky notes on a blankboard, engaging three distinct groups: those who have never used AI image generation, users with limited experience, and experts familiar with AI tools. Each group was split into two subgroups to compare their experiences using Discord Midjourney and the newly designed Midjourney Web Platform. The goal was to understand how visual aids and keyword assistance impacted their learning speed and overall platform understanding.

Research Question

1. How would you describe your first experience using both platforms for generating images?

2. Did the Style Reference Library and Clickable Models help you understand the outcome of your prompt better?

3. How did the suggested keywords for prompts impact your experience when trying to create an image?

4. Which platform (Discord Midjourney or Web) made you feel more in control of the image generation process and why?

5. How quickly did you feel you could navigate and learn to use the features of the Midjourney Web Platform compared to Discord?

   
   

   

My Main Insight

A way to streamline AI image generation workflow…

During my project research, I focused on improving user experience in AI image generation platforms. I discovered that using features like clickable models and a Style Reference Library helps users easily visualize potential outcomes. Adding real-time task monitoring also makes managing workflows smoother, minimizing interruptions and improving overall efficiency.

Users with access to a visual Style Reference Library and Clickable Models tend to experiment more with their prompts. On average, they apply the same prompt across different models and references 4.22 times, significantly higher than the 2.88 times for users without a visual panel. This data suggests that the visual interface encourages users to explore more creative possibilities, enhancing their engagement and interaction with the platform.

Testing + Improvement

3 Key Improvements in My Design

Based on feedback from 8 users and mentor reviews, I iterated my design over 2 weeks to address user needs and improve functionality, resulting in 3 key improvements:

Design System