The paper introduces HyperDreamBooth, a hypernetwork that efficiently generates personalized weights from a single image, enabling fast and high-quality face synthesis in various styles and contexts. It achieves personalization 25x faster than DreamBooth and yields a much smaller model.

https://arxiv.org/abs//2307.06949

YouTube: https://www.youtube.com/@ArxivPapers

PODCASTS:

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers