Dec 17, 2025 07:00:00

Apple unveils 'SHARP,' a method that can convert a single image into a 3D scene, which can be processed in less than one second on a standard GPU

Apple has announced a method called 'SHARP' that uses neural networks to generate highly accurate 3D scenes from a single image, and can run in under a second on a standard GPU.

[2512.10685] Sharp Monocular View Synthesis in Less Than a Second

https://arxiv.org/abs/2512.10685

SHARP
https://apple.github.io/ml-sharp/

GitHub - apple/ml-sharp: Sharp Monocular View Synthesis in Less Than a Second
https://github.com/apple/ml-sharp

Given an image, SHARP returns the parameters of a 3D Gaussian representation of the depicted scene via a single feedforward pass through a neural network, which can run in under a second on a standard GPU.

The 3D Gaussian representation output by SHARP is saved in the output folder as 3D Gaussian Splatting (3DGS) . 3DGS extracts features from images and videos, creates a collection of points called a point cloud, and averages the surrounding color information for each point using a Gaussian function. By placing colors in 3D space, it is possible to create a highly realistic 3D space.

The image above is the image input to SHARP, and the bottom is the 3D scene generated by SHARP, moved and viewed from a different angle. You can see that even though the angle is different from the original image, the details are still expressed in great detail.

Using a different image as input, the researchers generated a 3D scene. Compared to previous models, the Sharp model achieved 25-34% improvements in

LPIPS and 21-43% improvements in DISTS benchmark scores.

While the relationship with SHARP is unclear, Apple has already commercialized a 3D scene generation technology called '

Spatial Scenes .' Spatial Scenes will be available starting with iOS 26 , released in September 2025, and uses the Neural Engine, an AI-dedicated processor built into Apple devices, to generate three-dimensional images (spatial scenes) from 2D images.

You can see how a spatial scene is actually generated from a 2D photo in the video below.

I tried making photos three-dimensional using iOS 26's 'Spatial Scene' feature - YouTube


Open your iPhone's library and tap the icon in the top right corner of the image.

The image processing began, and the spatial scene was generated in just a few seconds. When you tilt the iPhone, the photo moves in three dimensions. It's notable that the sky behind the branches and leaves, which wasn't visible in the original image, is now depicted.