New Delhi: Apple has recently quietly unveiled one of its most impressive AI research projects yet, a model that can convert a single 2D image into a lifelike 3D scene in less than a second. The system, called SHARP, short for Sharp Monocular View Synthesis, can be performed in less than a second and has been detailed in a newly published research paper and made available publicly on GitHub. SHARP lets users take an ordinary image and view it from slightly different angles as if users were standing in the scene themselves. It doesn’t just fake depth with filters; it actually builds a 3D representation of the image that feels realistic when viewed from multiple perspectives.
This is the latest showcase of its growing expertise in advanced computer vision and AI research, fields that underpin everything from iPhone photography to the company’s work in augmented and virtual reality. The model analyses a single image and predicts what the scene might look like in three dimensions. It then constructs a lightweight 3D map made up of millions of tiny data points called Gaussians. Each Gaussian acts like a fuzzy bubble of colour and light positioned in space. When rendered together, these points create a realistic, dynamic scene that can be viewed from nearby angles, complete with smooth depth shifts and parallax effects.
SHARP isn’t meant to create fully explorable 3D worlds or fill in parts of an image that were never captured in the first place. It focuses on local viewpoint changes, meaning users can look at the scene from slightly different angles but not walk around it entirely. Apple has also acknowledged this limitation in its research paper, noting that the model is designed to maintain realism rather than hallucinate new details. The goal is believable depth, not fantasy reconstruction. By publishing code and research online, the company is inviting developers and academics to experiment, adapt, and build on top of its work, a rare move from Apple, which typically keeps its cutting-edge projects under wraps.
