A casual pet project that enables users to create new looks of themselves from a set of reference images.
Role
Product Designer
Type
Pet Project
Time spent so far
~ 2 weeks
How it started.
10 am (Whats app chat)
Kaushik :
Hey Kabil! I’ve got this fun idea around the new AI trend. Want to try it out as a side project?
Kabil :
Sounds cool! I’ve been wanting to play around with AI stuff anyway. What’s the idea?
Kaushik :
So, you know that Instagram trend where people use AI prompts to create those wild, creative photos? I was thinking — what if we build a simple tool for that?
Like, users can browse a bunch of sample images, pick one they like, upload their own photo, and get an AI-generated version. We’ll handle all the prompts and image generation behind the scenes.
Kabil :
That actually sounds pretty awesome. I tried something similar once with a photo of my car — it came out great. Let’s do it!
Kaushik :
Nice! Are you free this evening to chat more about it?
Kabil :
Yep, I’m in. Let’s hop on a call tonight.
Kaushik :
Perfect! Catch you this evening then.
We decided to build a MVP as fast as possible and role it out to users.
Started building some basic mockup screens needed to get started with development.
After the basic mockups were done, I shared it with Kaushik to start with the development.
I focused on getting the reference images and the prompts ready.
Kaushik focused on building the necessary front end backed and DB for the application.
Front End
React built with Cursor.
Backend and DB
We used Firebase.
Release MVP!
Shared the application with friends and family and few other people.
Tried some basic things in Instagram to increase traction.
I did usability testing with a handful of users.Below are the findings and improvement areas.
1.Hesitation during Sign in
Users sign in using their email ID, and we currently leverage Firebase Authentication for this process. However, when a user taps on the sign-in button, the URL briefly redirects to a Firebase domain. This sudden URL change can make users feel uncertain or suspicious, as it appears they’re being redirected to an unfamiliar site.
This has resulted in a significant trust issue that we need to address.
2.Slow loading of the landing page
Users land on a page displaying a gallery of reference images. Initially, when there were fewer than 10 images, the loading time was acceptable. However, as the number of images grew to around 30, the loading performance significantly declined, and users lost patience waiting for the images to load.
We need to optimize image loading. Potential improvements include implementing lazy loading and converting reference images in Firebase to more web-friendly formats, such as .webp.
3.Long time to generate image
It currently takes around 2–3 minutes to generate an image. Right now, we only display a simple loader, which often leads users to leave the browser and do other tasks.
We need a way to keep users engaged or at least inform them when their image is ready. Possible approaches include showing a progressive loader or sending a notification once the image is generated.
Since the image generation time depends on OpenAI’s processing, we don’t have direct control over it. We should explore the technical feasibility of these options and implement a suitable solution to enhance the user experience.
4.Handle errors gracefully
5.Make some UI changes like icons, colours etc.
What Next?
We haven’t made much progress since the usability testing.
While there are several areas that need improvement, we also need to think from a broader product perspective - things like
Marketing
Future scope
Expanding user base
Evaluating ROI.
Both of us have been quite occupied with other commitments, so we haven’t been able to dedicate enough time and effort to this project. However, we plan to revisit it soon and explore how we can grow it further. It’ll be interesting to see how it evolves.
Let's Work Together!
MAIL ME
kabilyogeshtamilselvan@gmail.com
FOLLOW ME
©2025 Kabilyogesh. Crafted personally with Framer.