Huhu Studio: 01 AI Virtual Try-On SaaS Platform For Enterprise Clients

Huhu Studio: 01 AI Virtual Try-On SaaS Platform For Enterprise Clients

OVERVIEW

Huhu Studio is a B2B SaaS platform that enables fashion brands to generate AI try-on images at scale. The platform supports bulk image generation, real-time project feedback loop, and an image gallery that allows clients to efficiently organize assets. By combining AI automation with thoughtful UX, Huhu Studio helps clients reduce manual overhead while maintaining brand consistency across visuals.

IMPACT

Increased AI image generation throughput by 50%.

Increased image delivery efficiency by 50%.

The system is actively using global fashion brands including Cider, Liverpool, and Landmark.

Actively serving global fashion brands including Cider, Liverpool, and Landmark, contributing to a total of 1M+ deal.

More Info ↗

👈 This web page was also designed by me!

MY ROLE

IC Product Designer

TEAM

2 Product Designers (incl. me)

1 Product Manager

1 UX Engineer

1 Front-End Engineer

1 Back-End Engineer

TIMELINE

Nov 2024 - Apr 2025

CONTEXT

Huhu AI helps enterprises generate AI virtual try-on images at scale.

The company was in the stage of pilot testing with global fashion brands such as Shein, Cider, and Liverpool, helping them generate AI fashion model and virtual try-on images tailored to each brand's specific SKU requirements and visual standards.

CHALLENGE

Excessive context switching slows monthly throughput.

However, each touchpoint requires switching between spreadsheets, AI image tools, and Figma—from organizing client feedback to generating and documenting images per SKU—leading to significant time lost in context switching and thus, non-scalable throughput.

PROBLEM STATEMENT

How might we help image operators streamline workflows and manage client data efficiently through an all-in-one SaaS platform?

How might we help image operators streamline workflows and manage client data efficiently through an all-in-one SaaS platform?

PROCESS

From MPV mapping to driving daily iterations.

My process was divided into two key phases: first, research and MVP mapping — digesting complex PRDs, identifying user needs, and outlining core user flows; second, rapid iteration — once the system has been implemented, I started working closely with developers to refine features and respond to evolving requirements. This approach ensured speed without compromising clarity or usability.

MY ROLE

Sole designer for Project Creation and Brand Management flows to lay the foundation for efficient service delivery.

I was responsible for the end-to-end design of two key user flows. The first enabled internal ops team members to seamlessly upload enterprise project requirements. The second allowed them to easily manage and review AI-generated fashion model images at a glance.

Project Creation Flow
Brand Management Flow
IMPACT

Designed and implemented the system in just 2 months, enabling Huhu to sign deals with larger scale.

Image Generation Throughput

+50%

+50%

Global brand deals secured

$1M+

$1M+

Internal DAU

30+

Internal DAU

30+

30+

DEEP DIVE - PART ONE

Submitting SKU-level Project Requirements With Ease

Submitting SKU-level Project Requirements With Ease

001

Project Input Via Spreadsheet Upload & Manual Input

Some large clients manage complex SKU requirements in spreadsheets, while SMBs often handle fewer SKUs per project. Supporting both spreadsheet uploads and manual input enabled us to meet the needs of both groups and integrate seamlessly into their workflows.

002

Manual Edit For More Sense Of Control

Clients' requirements often evolve after initial submission—such as adding new images or updating preferred models for specific SKUs. Providing manual edit options ensure that new information can be updated seamlessly.

Edit Text Fields
Edit Image Fields

To prevent errors from editing project fields, I designed two modals to clarify communicate limitations & permission. 👉

To prevent errors from editing project fields, I designed two modals to clarify communicate limitations & permission. 👇

To prevent errors from editing project fields, I designed two modals to clarify communicate limitations & permission. 👇

003

Bulk Edit Multiple Rows At Once

Some fields are often consistent across multiple SKUs. To save users time and reduce repetitive input, I introduced a bulk edit feature that lets them update shared fields all at once.

004

Clear Indication Of Erroneous Fields

To help users quickly identify and resolve input issues, I added a panel that summarizes the number + type of erroneous fields.

DESIGN DECISIONS

Designing in tandem with clients on both Content UX & usability.

Since we were building the platform while working with our initial customer, we had to learn by doing—adapting quickly to evolving client requirements was key. I iterated based on two main inputs: usability feedback to enhance the experience, and shifting client needs to stay aligned with their workflow.

001

How can table columns work smarter, not harder?

How can table columns work smarter, not harder?

Table Columns: From Specific to Generalized

Table Columns: From Specific to Generalized

The table columns were initially designed around one client’s specific needs. Over time, by analyzing more client behaviors, I identified common patterns. This allowed me to generalize the columns, refining the example spreadsheet format to better accommodate a wider range of clients requirements.

Mandatory vs. Optional Fields

Mandatory vs. Optional Fields

During testing, we discovered new columns like paired garment & shoes that could apply to other clients as well, so I added them as optional fields. Mandatory fields remain those will directly block progress if not specified at this stage.

002

Bulk Edit: How can we make it easier to find the right fields?

Bulk Edit: How can we make it easier to find the right fields?

All Fields, All At Once

All Fields, All At Once

I initially displayed all fields in bulk edit to give users full control, but it quickly became overwhelming and hard to scan.

💔 "I just wanna change one thing, but I have to scroll through all."

Focused And Searchable

Focused And Searchable

To improve usability, I switched to showing only the most frequently edited fields by default, with a search option to surface others as needed.

💚 "I like that it only shows the important stuff first!"

003

What’s more helpful: splitting or highlighting?

What’s more helpful: splitting or highlighting?

Split Into Two tables – Incomplete and Complete

Split Into Two tables – Incomplete and Complete

I initially separated the table into two sections to help users focus only on the one that needs editing. However, during testing, users found this disorienting. When a row was updated and moved to the other table, it disrupted their flow, especially in large projects.

💔 “I get lost when I fix something and it jumps away.”

Single Table With Visual Highlights For Incomplete Fields

Single Table With Visual Highlights For Incomplete Fields

I redesigned the experience into a single unified table. Instead of separating rows, I used visual highlights to draw attention to fields that still needed input.

💚 "I like seeing everything in one place—it feels more stable."

DEEP DIVE - PART TWO

Brand-Based Image Gallery Organization

Brand-Based Image Gallery Organization

001

Filter Design For Quick Sorting

By default, fashion model images in the gallery are organized using two key attributes: Model & Style which are the most commonly used by our users. For more tailored browsing, an advanced search option allows users to apply additional filters and specify values.

Default Filter

Dropdown with values, coupled with search bar + apply all option

Advanced Filter

All available attributes to filter

Selected attributes & values, with pagination and search bar

DESIGN DECISIONS

Iterating over concept hierarchy & usability.

I refined the brand management flow by prioritizing frequently used filters, introducing a clean attribute-value structure, and adding pagination and advanced search. These updates streamlined navigation and made asset discovery faster and more intuitive for users.

001

Style ID & Model ID: from hierarchy to harmony.

Style ID & Model ID: from hierarchy to harmony.

Top-Down Relationship

Top-Down Relationship

I initially designed the gallery with a top-down relationship, where Style acted as the parent category and Model was nested beneath it. This rigid hierarchy led to repetitive work when wanting to upload images.

💔 "Why can’t I just search by the model I need?."

Parallel Relationship

Parallel Relationship

I restructured the flow so that Style and Model now exist on equal footing, allowing users to filter by either attribute independently. This update made navigation more flexible and intuitive.

💚 "Much easier to to find / upload images now!"

002

How detailed should the filters be?

How detailed should the filters be?

Full Exposure

Full Exposure

I first designed the filter as a vertical panel with two commonly used attributes fully expanded. While it gave users quick access, it often felt visually cluttered.

💔 "I like how the values are displayed already, but it’s a bit much when I’m viewing all the model images on the side.”

Smart Reveal

Smart Reveal

To improve focus, I switched to dropdowns for the top two attributes. Users still get quick access, but with less noise. Additional filters are available within the dropdown for deeper refinement

💚 "Much cleaner and still easy to use.”

REFLECTION

My take designing a 0-1 B2B SaaS platform.

Start with the Right Blueprint
Defining the concept hierarchy—project, SKU, and task levels—was foundational. Without this clarity, I couldn’t determine what users should do or how the system should behave at each layer. Getting this right upfront shaped not just the interface, but also the entire user journey.
Misaligned Concepts = Major Rework

Misaligned Concepts = Major Rework
In the brand flow, an unclear split between “style” and “model” led to both backend and UI overhauls. It was a costly reminder that every design decision carries structural weight—especially in B2B SaaS where data models and interfaces are tightly linked.
Design for Now, Not Later
Early on, I designed for an ideal state that relied heavily on system data. In reality, we needed a simpler MVP flow with fewer dependencies. Starting lean and layering complexity over time would have saved us time and headaches.

Let’s connect &

make magic together!

©Tina Chen 2025 All Rights Reserved

Last Updated: 06/24/2025

Let’s connect &

make magic together!

©Tina Chen 2025 All Rights Reserved

Last Updated: 06/24/2025

Let’s connect &

make magic together!