Overview
AirBrush, the leading professional portrait retouching and editing software, aimed to incorporate AI technology into its existing solutions to drive market expansion and attract investment in the competitive AI landscape.
AI Professional Headshot is one of our GenAI initiatives, targeting professionals who require high-quality headshots for business profiles, LinkedIn, job applications, and more.
My Role
Design Lead
-
Led the design process from concept to launch
-
Orchestrated two key iterations
-
Collaborated with 2 PMs, 5 Developers, 1 Brand Designer
UX Research
-
Conducted prototype testing
-
Collected and analyzed statistics and user feedback
Standard Operating Procedures (SOP)
-
Organized design critique sessions
-
Established a localization management system
Problem Space
Obtaining professional headshots is often hindered by high costs, scheduling challenges, and the technical skills needed for quality results.
We aimed to create a solution to remove these barriers by providing an accessible, cost-effective, and easy-to-use alternative, allowing a broader audience to effortlessly create polished, professional images that elevate their digital presence without the usual hassle or expense.
$262.84
The average cost of a professional headshot in the US
3-5 Days
The average turnaround time for headshots at photography studios
14 Times
LinkedIn profiles with professional headshots receive 14 times more views than those without
Key Features
A web-based AI image generator designed to help customers effortlessly create impressive professional photos using facial data.
Generate AI Face Model
Users upload 15 photos to create a reusable digital face model for generating multiple professional headshot sets.
The interface provides immediate visual feedback on each photo's suitability, ensuring high-quality input for accurate results.
Select Styles
Choose from various professional headshot styles tailored to suit different business and social profiles.
Users can experiment with different styles, such as formal, casual, or creative, and preview how each look fits their personal brand. This feature provides flexibility in establishing the right tone for different platforms or career needs.
Multiple
Candidates
Due to AI's generative variability, we provide several headshot options within the same style, enhancing the likelihood of a satisfying result.
By offering multiple variations, users can select the image that best aligns with their preferences.
Retouch
Considering the imperfection of AI, we provide an editor for users to further adjust details such as makeup, hair color, skin tone, and more. This ensures that the final result better aligns with their expectations.
Business Context
Background
AI Race In 2022​
-
ChatGPT was launched on November 30, 2022. It surpassed 1 million users just five days after its launch.

-

In 2022 alone, 524 AI startups were founded in the U.S., attracting $47 billion funding. 


-
160 newly-founded AI startups in China received $71 million each on average.
Strengths & Opportunities
AirBrush
AirBrush has established itself as the leading professional portrait retouching and editing software.
The integration of AI technology presents a significant opportunity to enhance our solutions, expand market share, and attract venture capital.
Business Question
How do we strategically incorporate AI technology into our current solutions to drive market expansion and secure investment opportunities?
Beta Objective
Rapidly launch a fully functional Beta within one month, enabling us to quickly test and validate its market performance.
User and Product
Landscape
Before crafting the interface, I made sure to understand:
Target User
Professionals who require high-quality headshots.
​
Use Case
​Across various media, including team pages, social and LinkedIn profiles, business cards, pamphlets, brochures, email signatures, etc.
​
Pain Points
Common pain points people face when getting professional headshots in a photography studio:
-
Time-consuming scheduling
-
High costs
-
Limited accessibility
Beta
Design Principles
According to the business plan, the AI Professional Headshot would become a part of the AI Creative Set (Including AI Portrait, AI Avatar, AI Professional Headshot, and AI Face Model Generator).
I established three design principles to ensure alignment with our design and product goal:
Fully Integrated
Seamlessly incorporate this tool into the AI creative set for a holistic experience
Consistency
Align with the design patterns of other AI tools for a cohesive user experience
Efficient Development
Aim for launch within a month; reduce engineering overhead
AI Portrait
AI Professional Headshot (Beta)
Beta Design
Layout
Follow the Patterns of the AI Creative Set​
I designed the AI Portrait two months ago, and it was launched by the time I started working on the AI Professional Headshot.

To maintain consistency and achieve efficient development, I followed the same design patterns and utilized the same design system elements as the AI Portrait.
This not only ensured a cohesive user experience across our AI Creative Set but also streamlined the development process, allowing us to launch quickly.
Interactions
Create a Holistic Experience​
We sell the AI Creative Set as a whole, which includes three AI tools and a face model generator.
To ensure these tools are fully integrated, I designed an experience that allows users to use the face models they create in the generator to generate photos across all AI tools.
I aimed to provide a holistic user experience, allowing users to maximize the utility of their face models across different creative applications.
Beta Result
Business Result
Top Performer
AI Professional Headshot became the top performer among the three AI tools in the AI Creative Set.
Analytics
Low Completion Rate
Over 30% of users who visited the AI Professional Headshot Generation page did not complete the task. Plus, many of them got stuck on the first step without moving forward.
Assumption
The navigation is not intuitive for users. Users don't know how to proceed to the next step.
What We Needed to Do Next
We needed to dig deeper and pinpoint exactly where users were having trouble so that we could make the tool easier to use and increase completion rates.
User Feedback
Low Satisfaction
Additionally, many users expressed dissatisfaction with the outcome and submitted negative feedback through customer support.
Here are some specific points they raised:


-
The generated photos didn't resemble the users.
-
The photos appeared unrealistic and cartoon-like.
-
Ethnicity representation was inaccurate.
-
Religious attire was not represented in the final images.

1st Iteration Objective
Ensure users complete the flow effortlessly and clearly understand the product's value.
Strategy Pivot
Independent Tool
Based on market performance, we decided to pivot our strategy.
We started marketing and selling the AI Professional Headshot as an independent tool, rather than part of a bundle.
Additionally, we changed the SKUs to one-time purchase plans.
1st Iteration
Design Principles
After analyzing the user feedback and identifying the key issues, I established three new design principles for this iteration:
Seamless Task Flow
Enable seamless task completion in a single flow
Intuitive Interaction
Create intuitive user experiences
Enhanced Expectation-aligned Output
Ensure the delivery of ideal, high-quality headshots
1st Iteration Design
Task Flow
I redesigned the task flow, enabling task completion in one tool, simplifying it from the initial complex process to a more streamlined one, making it much easier for users to navigate and complete.
Before
Redesigned
Interactions
Create a Step-by-Step Guided Experience​
I created a step-by-step guided experience for users. I aimed to simplify the process by breaking it down into manageable steps.
STEP 1 Select Styles
STEP 2 Upload Photos
Ask users to provide their ethnicity information to help improve the AI output.
STEP 3 Preview & Download
Generate multiple photos for each style. This increased user satisfaction by giving them more options to choose from, ensuring they found a headshot that truly met their needs.
1st Iteration Result
Analytics
Improve Completion Rate from 65% to 91%
More than 90% of users completed the task.
Look into the Feedback
Numerous complaints about photo output
“These photos don’t look like me.”
How & Why?
I realized that we need to delve into the root to provide more instructive suggestions to our dev team for improvement.
What I did Next
Implemented a feedback form to gather specifics.
Delved into the Root
Why "these photos don't look like me?"
Through the data collected from the feedback form, I discovered three key insights:
-
Hair & Beard Mismatch
-
Hair & Eye Color Incorrectness
-
Skin Tone Inaccuracy
2nd Iteration Objective
Optimize the user experience and allow users to achieve their desired results.
Design for AI
AI-Human Communication
Throughout this journey, I kept asking myself: how is designing AI tools different from other tools? What makes designing for AI special?
By reflecting on my own takes during this journey and researching other AI design theories online, I started to see that the user experience between Humans and AI is fundamentally communication.
Designing for AI means setting up the right conditions for great communication in a dynamic human-machine relationship.
I found that: 

-
Users provide input data for AI to learn and want maximum benefit from minimal effort.

-
AI is like a black box, delivering both expected and unexpected results. Therefore as a designer, it’s important to think about how much control should be given to users over these outputs.
2nd Iteration
Design Principles
Based on this Human-AI communication framework, I found three aspects that we can address to achieve Expectation-aligned Output: Input, Output, and Algorithm.
For this iteration, I defined two design principles:
Clear Guidance
Assist users in providing accurate and necessary information
Provide User Control
Empower users with controls to customize AI outputs, enabling modifications to meet their expectations for AI headshots
2nd Iteration Design
Redesign
Aimed at Achieving Desired Results​
Focusing on improving the design to better meet user expectations, I reviewed all steps and made some changes.
Redesign 01 Input of Ethnicity
I got good news from the engineers: our AI Algorithm has been improved. Now, our AI model has enhanced the accuracy of identifying ethnicity.

Previously, in the MVP version, we didn’t ask for ethnicity inputs but received negative feedback. As a result, we had to ask users to provide their ethnicity information to help improve the AI output.
Now that technology has evolved, we can eliminate this input and minimize user effort in this iteration.
Additionally, I applied gender tags to clothing styles without asking users to specify their gender, thus avoiding ethical sensitivity risks.
Previous
Redesigned
Redesign 02 Input of Styles
Initially, I proposed showing example photos in the MVP version but faced pushback from the PM and engineers. They highlighted that displaying all combinations of options (colors, outfits, backgrounds, genders) would require significant storage and workload. To address this, I simplified the design by using drop-downs for text-based info.

In this iteration, I reduced the number of options based on user data and feedback, keeping only the most popular choices. This allowed us to include example photos while selecting styles, making it more visually straightforward.
​
Previous
Redesigned
Redesign 03 Guide of Photo Upload
Previously, the guidance was presented in a tooltip that only appeared when users hovered over an icon, which could be easily overlooked. 

I moved the guidance directly onto the page, making it more visible upfront. Additionally, I simplified the content from sentences to more concise terms for better readability, ensuring users understand the requirements quickly and easily.
Previous
Redesigned
New Feature Edit Photo
Since AI is not perfect, I decided to allow users to retouch their headshots to ensure they achieve optimal outcomes.

To provide users with control over the final product, I introduced an "Edit Photo" option. Users can now make adjustments to eye color, skin tone, and hair color. This feature empowers users to fine-tune their headshots to better match their preferences and needs, ensuring higher satisfaction with the final images.
Final Results & Impacts
Analytics
Improve User Experience
-
Task Completion Rate
June (Beta): 65%
October (V2.0): 91%
​​
-
Photo Download Rate
June (Beta): 56%

August (V1.0): 76%
October (V.2.0): 82%​
Improve User Satisfaction
-
Net Promoter Score
August (V1.0): NPS of 35
October (V2.0): NPS of 50
​​
-
Refund Rate
June: 11% refund rate
October: 3% refund rate​
Enhanced Monetization
-
Paid Users
June: 70

August: 4000
October: 7000
​​
-
GMV (Gross Merchandise Value)
0 - October over $60k
Public Reputation
Ranked #2 Product of the Day on Product Hunt
Received
Enterprise Inquires
We've received some fantastic feedback beyond just the numbers.
Several small and medium businesses reached out, asking for a solution to help their employees generate professional headshots.
This led me to design a B2B enterprise solution tailored to their needs. It was incredibly rewarding to see how our tool could be adapted and embraced by different markets and user groups!
Integrated the Feature into the AirBrush Mobile App
Due to the significant success of the AI Professional Headshot, we decided to integrate this feature into the AirBrush mobile app, which has a user base of 30 million.
This provided users with an accessible and lightweight solution for creating professional headshots. Within just 3 months, it achieved over 10 million generated outputs.
Project Takeaways
Adaptability
Embrace an adaptable mindset, anticipating various shifts such as technological advancements, market dynamics, and product positioning strategies.
​
Designing in New Domains
When faced with limited benchmarking options, deconstruct user needs, reanalyze each interaction, and strike a balance between introducing new interactions and leveraging universal patterns.
​
Effective Communication
Effective communication entails comprehending the AI's underlying logic and reshaping the product through the lens of user perspectives.