Sharp-It: A Multi-view to Multi-view Diffusion Model for 3D Synthesis and Manipulation

Supplementary Material

Click below to view our interactive results, comparisons, and ablations:

Sharp-It introduces a novel multi-view to multi-view diffusion model that refines geometric details and textures of 3D objects generated by low-quality generative models. This supplementary material shows the model's effectiveness in enhancing 3D content creation by leveraging a 3D-consistent multi-view set, enabling high-quality synthesis, editing, and controlled generation.

Comparisons to Other Methods
Comparisons to Zero123++
Normal map examples


Our Results

Sample results from our framework over all three presented tasks.

Rainbow Chairs

Input
Rainbow Leather Chair
Rainbow Velvet Chesterfield Chair
Rainbow Minecraft Chair

Knights

Input
No Prompt
Single Source of Light
Full Model

A glass table with intricate gold legs

Input
A glass table with intricate gold legs

A stained-glass art-deco lamp

Input
A stained-glass art-deco lamp

A dragon

Input
A Dragon

A gold lamp

Input
A gold lamp

A Santa lamp

Input
A Santa lamp

A turquoise beetle car

Input
A turquoise beetle car

A blue SUV

Input
A blue SUV

A wooden tower

Input
A wooden tower

A leopard print leather chair

Input
A leopard print leather chair

A suede leather jewelry box

Input
A suede leather jewelry box

A red velvet chesterfield chair

Input
A red velvet chesterfield chair

A cargo spaceship

Input
A cargo spaceship

Additional Results

A fireplace with logs

Input
A fireplace with logs

A tiled wooden hut

Input
A tiled wooden hut

A yellow Vespa Scooter

Input
A yellow Vespa Scooter

A rainbow spaceship

Input
A rainbow spaceship