![Stability.ai unveils Stable Point Aware 3D for Real-Time 3D object creation from images](https://d4.alternativeto.net/Ab_WxeH5ta_v0nd55RWVKAuZSw9zE98yceChPGUIcso/rs:fill:760:380:0/g:ce:0:0/YWJzOi8vZGlzdC9jb250ZW50LzE3MzY0NDczNTA3NjUucG5n.png)
Stability.ai unveils Stable Point Aware 3D for Real-Time 3D object creation from images
Stability.ai, renowned for its Stable Diffusion technology, has launched a new tool called Stable Point Aware 3D (SPAR3D), offering real-time 3D object creation and editing, and allowing users to generate and modify 3D objects from a single image in less than a second. SPAR3D integrates precise point cloud sampling with advanced mesh generation, and employs a dual-stage processing approach, generating a detailed point cloud using a point diffusion model and reconstructing high-resolution geometry, texture, and lighting with a triplane transformer.
SPAR3D provides comprehensive editing capabilities, enabling users to manipulate point clouds directly by deleting, duplicating, stretching, recoloring, or adding features to customize their 3D models. It offers complete structure prediction, delivering accurate geometry and detailed 360-degree views, including typically hidden areas. The tool's processing speed is impressive, with final meshes generated in just 0.3 seconds and detailed 3D meshes from a single image in 0.7 seconds.
SPAR3D is free for commercial and non-commercial use under the Stability AI Community License, with weights on Hugging Face, code on GitHub, and API access via the Stability AI Developer Platform. Enterprise licensing is available for larger businesses, offering scalability for organizations with annual revenue exceeding $1M.
SPAR3D sounds like a game-changer for 3D content creation! The speed and precision it offers, especially from just a single image, are incredible. Excited to see how this impacts industries like gaming, design, and AR/VR development!
The rendered models seem to be pretty rough but the generated texture (and UV projection) seems decent. There about a million teapot models on internet so they've chosen on of the most simple cases, and more complex models are giving moderate results (even bad when they're not clearly separated), but this could be interesting to generate some parallax effect on 2D images.