-
Screenshots of our app
-
Screenshots generated by our app
Inspiration
Every few months or so, smartphone users such as ourselves get tired of staring at the same wallpaper every time we unlock our phones. Searching the web for a cool wallpaper that matches your aesthetic and personality can be a tedious and inconvenient process.
What it does
We thought it would be great if new and unique wallpapers could be generated from a simple description. This is why we combined our efforts to create WallpaperAI; a user-friendly application that uses machine learning (ML) to create images.
How we built it
Our adaptive full stack mobile application utilizes an open-source machine learning model to deliver users blazing fast custom-generated wallpaper-sized images. The machine learning model that we are using (called Stable Diffusion) was just released last month and relies on decades of research in the field. This model has 2 billion parameters, meaning that it has an extremely large training data set, enabling this advanced image generation. This model has been extremely compressed and enables the almost instant image generation offered by our app. In order to accomplish the UI, we leveraged the latest tools available; Combining Ionic with React to deliver an appealing but easy-to-use UI on the front-end and creating a custom Node.js backend for flawless integration. Once the web app was finished, we used Ionic Capacitor to convert it into a mobile version.
Challenges we ran into
We encountered two main challenges: (1) Finding an open-source machine learning model which allowed us to integrate it into our back-end, and (2) actually integrating it into the back-end once we found one. Once we began the integration process, we ran into a variety of cors errors which limited us from getting an output from the server. Some of us also never used the Ionic and React frameworks and struggled to find educational material that matched our vision.
What we learned
After much trial and error, we learned how to integrate open-source APIs into web apps with Node.js and became more skilled in React along the way.
Accomplishments that we're proud of
When finishing up the project, someone from a different team came up to us, curious about what we had been working on. We explained it, and she was eager to try her own wallpaper description. We are proud to say that she was very thrilled with the AI-generated wallpaper and asked us to send it to her right away.
What's next for WallpaperAI
We aim to distribute the app globally so that millions of users can access this cutting-edge technology and download it through the app store with ease. The days of scouring the internet for wallpapers are over!
Log in or sign up for Devpost to join the conversation.