Apple is pushing its artificial intelligence research forward at a fast clip, and its latest breakthrough could change how quickly we create and interact with 3D content. In a newly published research study called “Sharp Monocular View Synthesis in Less Than a Second,” Apple details an AI model named SHARP that can generate a photorealistic 3D scene from a single 2D image—finishing the job in under a second.
That speed is the headline feature. According to the study, SHARP can produce its 3D result “less than a second on a standard GPU” using a single feedforward pass through a neural network. In simple terms, it doesn’t need a slow, multi-step pipeline to get from a flat picture to a realistic 3D view. It does it in one fast sweep.
So what does SHARP actually generate? The model creates a 3D Gaussian representation of the scene, which can then be rendered in real time to produce high-resolution, photorealistic images from nearby viewing angles. Apple also notes that the representation is metric and includes absolute scale, which matters for realistic camera motion. That means the 3D scene isn’t just an artistic approximation—it’s built in a way that supports believable, consistent movement as the viewpoint shifts.
To understand why this is significant, it helps to know what 3D Gaussian Splatting is. It’s a technique for building realistic 3D scenes by representing them as enormous collections of tiny, colored “splats” (think miniature blobs of color and depth). Traditionally, creating a high-quality scene this way requires multiple 2D images of the same subject taken from different angles, so the system can reconstruct depth and detail more accurately.
SHARP aims to remove that requirement. Instead of needing many photos, it predicts key properties like depth and color from just one image, then generates a complete photorealistic 3D Gaussian scene representation—quickly enough to feel nearly instant.
Apple’s paper emphasizes that SHARP estimates what the scene should look like from “nearby viewpoints,” essentially inferring how the view changes as you move the camera slightly around the original image. That ability is a big deal for applications that benefit from fast novel view synthesis, such as immersive media, rapid 3D visualization, creative tools, and potentially future AR and VR workflows.
Apple has also made SHARP available for people who want to experiment with it themselves, offering access through its official project page and accompanying research paper—without needing to pay to test the model.
