The model utilizes a novel training approach that "unlocks the potential of the diffusion prior".
Researchers Junzhe Zhu and Peiye Zhuang have introduced HiFA, a novel high-fidelity text-to-3D method with advanced diffusion guidance. The method utilizes the diffusion prior and a novel training approach to maximize the potential of the diffusion prior, enabling advanced photorealism and improved multi-view consistency. Additionally, the method applies auxiliary depth supervision for NeRF-rendered images and regularizes the density field of NeRFs, enhancing 3D geometry representation.
And here are some 3D meshes the team generated using the proposed method:
Prompt: A baby bunny sitting on top of a stack of pancakes
Prompt: A beautifully carved wooden knight chess piece
Prompt: A highly detailed stone bust of Theodoros Kolokotronis
Prompt: Small saguaro cactus planted in a clay pot
Prompt: Head of thanos
Prompt: Iron throne from game of thrones
Prompt: A ladybug
Prompt: A high pile of chocolate chip cookies
Click here to learn more and check out more results. Also, don't forget to join our 80 Level Talent platform and our Telegram channel, follow us on Instagram and Twitter, where we share breakdowns, the latest news, awesome artworks, and more.