Input requirements

One of the challenges of training an AI model to generate avatars is the need for high-quality input data. In many cases, this requires input images that are clear and free from noise or artifacts, and that only contain the face of the subject without any accessories or obstructions.

However, by being able to train on input data that is less strict and includes accessories, Cavatar is able to work with a wider range of input images and produce more varied and interesting avatars. This can be particularly useful in applications where users may want to create avatars that include accessories or other elements that are important to them.