Season3 Project: AI Portrait

Team Introduction

Setteu - A product enthusiast who believes that the best products come from a team that is obsessed with the essence of the given problem and trust

Pluupy - An entrepreneur and developer who wants to create a global AI company

Woody - A curious problem solver

No more fake AI photos! Get your ID photos with IDSnap.ai now😎

“IDSnap.ai: Snap Your ID Photo, Simplified.”

When you need a job or need an ID photo, it's not easy to find a photo studio, make a reservation, take a photo, and adjust the photo every time. Moreover, the AI-generated profile photos these days may look pretty and handsome, but they often feel awkward as they don't seem to be truly yours. We introduce IDSnap.io, an easy and convenient AI ID photo service for busy modern people!

Have you heard about Snow AI profile photos?

You might feel embarrassed by your excessively pretty and handsome face.

But, aren't they very similar?

Currently, the top 3 AI ID photo services are Snow, Carrot, and Pikachu Studio. Snow is highly polished, Carrot has a soft sensibility, and Pikachu Studio generates results until you like them, but they all have the common limitation of not being controllable in the desired way. Our service is simple, fast, and optimizes the results according to the desired theme. You can create the desired ID photo in a few minutes and access it anytime, anywhere. You can receive the result right in front of your house with just one click, without being constrained by complex poses or lighting, just like in a photo studio.

So, we created it.

You don't have to go to a cumbersome photo studio anymore. It's a service that creates ID photos that look like you using AI.

IDSnap.io service ensures that you no longer feel 'strange' in your photos. With the advancement of generative AI technology, our AI creates ID photos that reflect your unique facial features most naturally. Now, you can easily get an ID photo that embodies your own personality, from your smile to your expression.

Key Features of IDSnap.ai

• Select a theme and you're done with just one image!

  • We provide 4 themes: women, men, neat, and casual hairstyles.

• You can create it anytime, anywhere by accessing the website!

  • It's easy to use with a responsive website, regardless of whether you're using a laptop or a phone.

• No need to wait; you'll be notified as soon as it's created!

  • You don't have to wait while the AI service creates the result; just enter your email and you'll be notified as soon as it's created.

[Input Image]

[Demo Video]

[Result ID Photo]

How to use IDSnap.ai

Step 1: Choose a theme

Step 2: Upload a photo taken as frontally as possible

Step 3: Enter your email and click the generate button

Step 4: Wait for the result

Development Process

1. Defining the problem and key features

   1. It started with the opinion that existing AI profile generation services such as Snow and Carrot have too much difference from the actual face, and aimed to solve the problem of how to maintain your own face while changing clothes or background according to the desired theme.

   2. We provide an AI studio service that can replace urgent ID photos taken at subway unmanned instant photo booths.

   3. We aim to satisfy users who want better quality than offline photo studios, while producing photos as neatly and expressing their own faces as well regardless of the angle, background, or lighting.

2. Product definition and requirements

   1. Devices and Environment

      1. Frontend (FE): Mobile web as main, implemented using Bubble (https://bubble.io/) tool

      2. Backend (BE): Store results in Amazon S3 bucket and notify users when generating ID photos!

   2. User flow

      • Page access → Upload photos → AI image generation → Check AI image result → Regenerate or complete

      • Expected flow after completion

        1. Download or share photo files (personal email or SNS integration)

        2. Photo printing (offline delivery paid service)

        3. View other users' photos (internal SNS service)

   3. Wireframe

      

3. Development of AI image generation

   1. We ask users for a front-facing photo, provide the same level of effect as a conventional photo studio for the facial part, and provide the rest (hair, clothes, background, expression) according to the theme

   2. Aim to provide 2 themes for men and women respectively

   3. IDSnap.ai Image Generation QA

Before: We tested applying the LoRA model through Web UI, but faced problems such as facial deformation, dressing according to the desired theme, and changing poses.

After: We improved performance by first selecting sample images according to the desired theme and changing only the area corresponding to the face! By combining tools like img2img + controlnet + faceswaplab, we found a way to maintain the format like actual ID photos while representing our own faces well, providing random results by varying the degree of the denoising parameter.

What I learned from the project

• All members gained a brief understanding of what AI is and how it is applied and developed for actual services. We were able to study various people and current technologies and gain a lot of know-how on how to proceed with the project.

• Above all, we deeply studied recent generative AI, learned how to use new tools, and applied them.

• As a service based on AI, we pondered on how to deal with the time it takes to generate the result, and it was also good to be able to create results.

• As we proceeded with the project as a team, we had the opportunity to think about how to communicate with team members effectively and learn how to schedule and discuss technical matters when collaborating.

Team Members' Reviews

• Setteu (PM): Schedule management, planning, and wireframe

  • I was very worried about the completion of the product overall, but I was pleased that it was completed in line with the Proof-Of-Concept (POC) specs.

  • I was disappointed that I couldn't participate much in the actual image generation AI part because I couldn't use Stable Diffusion smoothly on the devices I had.

• Pluupy (Developer): BE, AI image generation API development, setting up email sending environment

  • I had a lot of fun with the AI project for 4 weeks and used a technology that I didn't know well, generative images. I feel proud that the performance seems to be coming out to some extent.

  • It was an opportunity to learn a lot about the technical aspects related to generative AI-based image creation and to use many latest AI tools such as Web UI and LoRA.

• Woody (Developer): FE, wireframe development, image generation and email sending API response

  • I would like to emphasize that I was able to learn various uses of Stable Diffusion. It was also very good that the results came out to the desired level.