Inspiration
We found that text-to-image models, such as DALL-E, struggle to produce a consistent theme unless applied with intentional prompt engineering, which can be difficult for inexperienced users.
What it does
Rembrandt provides an interface for generating images with cohesive branding and aesthetic to a set of reference images. We envision Rembrandt to be a tool for producing professional client-facing presentations.
How we built it
We used JavaScript with React for the frontend. For the backend, we used FastAPI. We used OpenAI's Vision and Image Generation APIs to analyze and generate images based on user input. We then used AWS EC2 to host the backend and Github Pages to host the frontend.
Challenges we ran into
Many team members were unfamiliar with React, so we spent a lot of time learning the framework. We also encountered issues configuring HTTPS with EC2.
Accomplishments that we're proud of
We built out a sleek and easy-to-use interface using a new framework, and we fully deployed our project. We experimented and learned a lot about prompt engineering throughout the process, and we were able to develop our idea beyond our initial plan.
What we learned
Our group learned how to interact with OpenAI's APIs and use React to display dynamic web pages that queried our backend.
What's next for Rembrandt
Rembrandt can evolve into a tool where organizations can create accounts with pre-set themes for generating additional images. Rembrandt could also be integrated as a browser extension to allow for quick access to these images when creating slideshows and documentation.
Log in or sign up for Devpost to join the conversation.