Caricature AI

CaricatureAI was a project I developed to generate personalized caricatures.
The core technology leveraged Dreambooth, a cutting-edge image generation model, to create unique caricatures from a set of 5-10 input images. The MVP was simple and straightforward. After payment, users would upload their images, and once the generation process was complete, they could download a zip archive containing their personalized caricatures. I integrated the Dreambooth Replicate API into Celery async tasks running on a Hetzner VPS. The user interface was built using SvelteKit, while the backend utilized Django with Django Rest Framework (DRF). Despite exploring various distribution channels, the Customer Acquisition Cost (CAC) proved too high, thus making the project unprofitable. Consequently, I made the decision to discontinue it.