Inspiration

The idea of our project was to give people a tool that can quickly and easily generate high-quality wallpapers and other images using AI. With recent advancements in AI technology, many models have been trained to generate realistic and visually pleasing images that can be used for a variety of purposes. Although this technology is widely used, it is often difficult to get exactly what the user wants. We aimed to develop a user-friendly tool that could leverage it to generate unique and personalized experiences. To this day, the digital art scene is a growing place, and we believe that AI art is here to stay.

What it does

Given an input image by the user, Shakr.ai produces a beautiful image of similar style to the provided image.

How we built it

By drawing on the powerful tools of Stable Diffusion and its Image to Text counterpart, CLIP, we combined the two into a single pipeline, generating images similar to the input. In came our wonderful UI Team (Consisting of two people) to bring this powerful tool into a user friendly interface that anyone could interact with.

Challenges we ran into

Due to a lack of compatible computing power, where not one of our members had a discrete GPU to perform calculations in parallel, we were forced to run CLIP on a designated CPU Fast mode, where speed was prioritized heavily over quality of text generation. This then translated over to our Stable Diffusion implementation, where image quality was severely hindered by our machines.

Accomplishments that we're proud of

We were able to generate a stunning website in just 24 hours, while also being able to make use of powerful tools that we were unfamiliar with, and still creating a great product.

What we learned

During this 24 hour period, we were able to gain skills in AI usage, Generative AI, and Web design with vanilla HTML+CSS+JS. We also were able to interact with HuggingFace, Kaggle, and Github in our quest for a image to image model.

What's next for Shakr.ai

Generating a single image from another to use as a wallpaper was just the beginning. Next, we plan to take this to the next level, and generate a live wallpaper from a single image through frame extrapolation, giving people the dynamic backgrounds they wish for on the images they love.

Built With

Share this project:

Updates