Apple Unveils AI Tool That Converts 2D Images Into 3D in Under a Second

Apple has introduced SHARP, an AI research tool that turns a single 2D photo into a photorealistic 3D scene in under one second. SHARP creates a 3D representation you can view from nearby angles, so the image gains realistic depth and parallax.

Apple published SHARP through its Machine Learning Research team and released the project as open source, including code and model weights.

What SHARP does

SHARP stands for Single-image High-Accuracy Real-time Parallax. It takes one still image and produces a 3D scene representation in a single forward pass, which keeps the process fast and interactive.

After SHARP builds the scene, it can render new nearby viewpoints in real time. You can “look around” the photo slightly and see natural parallax, similar to how depth works in the real world.

How the technology works

SHARP predicts a 3D Gaussian scene representation, often associated with Gaussian splatting-style rendering. This representation stores the scene as many 3D elements, which helps SHARP render clean details and fine structures during viewpoint changes.

Apple also describes the output as metric, meaning the reconstruction preserves real-world scale. That detail matters because it supports more realistic camera motion for AR and VR-like viewing.

Performance and benchmarks

Apple reports that SHARP completes scene generation in less than one second on a standard GPU. The resulting representation can render nearby views at 100+ FPS on a standard GPU, which supports smooth interaction.

In its paper, Apple also reports large improvements on common perceptual quality metrics (including LPIPS and DISTS) compared to earlier approaches, alongside major reductions in synthesis time.

Open-source release

Apple released SHARP as an open-source project, including model weights and code. Developers and researchers can reproduce the results, test the model on their own images, and build tools on top of the pipeline.

Why this matters

Fast single-image 3D reconstruction opens the door to more interactive “spatial” photo experiences without multi-shot capture. SHARP also fits neatly into workflows that need quick depth-aware rendering, including prototyping for AR and VR scenes.

Apple has not announced a consumer feature powered by SHARP, but the research shows Apple can now generate convincing 3D scenes from a single photo at real-time speeds.

What’s next

SHARP currently lives as a research release, not a product feature. If Apple ships it in an app or OS feature later, expect the first uses to focus on interactive photo viewing, spatial media creation, and AR-ready scene depth.