Designing a text to image AI generation tool
This is a design exercise that I worked on while interviewing at an AI image generation company, and it provides a good look into my approach to tackling design challenges like this. The assignment for this project was to design a text to image generation product for both producers and consumers of generated imagery. I spent approximately 6-8 hours on this project. Everything following this introduction is what I submitted as my solution without any additional editing.
Goal
Design the image generation workflow for a text to image interface, considering the experience of both content creators and consumers.
Questions
These are a list of initial questions that I might ask if this were a live design exercise. Answers to these might lead to further questions.
- What are the business goals or objectives that this design needs to achieve?
- What are the KPIs of this service?
- Engagement rate, revenue per user, CTR, DAU, session duration, etc
- Who are the target users and what are their key needs and pain points?
- Are there any key technical constraints?
Assumptions
These assumptions consists of my own answers to the above questions and provide a framing to the solution
- Users are non-creatives (they are not professional illustrators, graphic designers, etc)
- Goals:
- Ease of use: The interface should be simple to navigate and understand
- Guided creativity: With non-creatives being the target user base, users require prompts and assistance in the creation process
- Community building: This has been identified as a key driver to engagement with the platform
- Mobile-first
User profiles
These profiles consist of several potential target users of this service and were constructed based on assumptions. They are not exhaustive by any means, but aid in providing some user goals and needs that help guide and focus the design process.
Content creators
Creators who require visuals for videos, blogs, and social media posts
Needs
- Ability personalize output, especially with typography
- Ability to quickly generate content that aligns with existing branding
- Customization options
Goals
- To generate visuals that align with their personal brand and are attention grabbing on social media
- To aid in generating visuals for merchandising
Hobbyists
Users who require visuals for personal crafting projects or to create personalize gifts for friends and family
Needs
- Creative inspiration
- Integration with existing crating tools and software
- High-quality images (print-quality)
- Robust organization of past output
Goals
- To create personalized images for crafting projects, such as greeting cards, cake toppers, posters, etc.
- To quickly explore and experiment with different styles and designs from inspiration and brainstorming
Corporate professionals
Corporate employees who work in professional environments in HR, finance, marketing, etc but do not have no visual or graphic design expertise.
Needs
- Easy to integrate into existing productivity tools
- Defined styles and consistent themes
- Help with layouts
Goals
- To quickly add visuals to existing decks and reports
- To generate images that are contextually relevant their content
User stories
For the purposes of this exercise I’ve created a single user flow for a single target user.
Target user:
Content creator
User flow:
Generating a new asset for a newsletter that already has an established brand
User stories:
- As a content creator, I can group my work into collections and view these collections by component type and style
- As a content creator, I can easily generate new assets that aligns with my existing branding
- As a content creator, I can easily remix and refine my assets to ensure they align properly with my brand
Solution overview
By selecting these 3 stories, I wanted to focus my design efforts on 3 areas that are important to content creators who may not have any formal experience in visual or graphic design:
- Allow users to group their output into collections that dynamically organize that content into types while highlighting commonalities
- Encourage users to build off their existing collections when generating new content
- Provide users with the ability to fine tune their output with more precision.
These features allow users to ensure that their output follows a consistent brand and allows us to inform and educate users about the graphic design building blocks of that brand, such as colors, typography, and illustration style. By encouraging users to build off their existing collections and giving them tools to fine tune their creations with more precision, Ideogram can become a much more powerful assistant for creators who do not have a visual or graphic design background.
Screens
User story #1
As a content creator, I can group my work into collections and view these collections by component type and style
Screen 1: My work page Screen 2: Collections tab on My work page
Screen 3: All assets tab of My newsletter collection page Screen 4: By component tab of My newsletter collection page
User story #2
As a content creator, I can easily generate new assets that align with my existing branding
Screen 1: Image generation sheet Screen 2: Add to a collection tab of Image generation sheet
User story #3
As a content creator, I can easily remix and refine my assets to ensure that they align with my brand
Screen 1: Remix tab in Keep going flow Screen 2: Refine tab in Keep going flow
