Apple's SHARP running in the browser via ONNX runtime web

Hi HN, author here. SHARP is Apple's recent single-image 3D Gaussian splatting model (https://arxiv.org/abs/2512.10685). Their reference code is PyTorch + a pretty heavy pipeline; I wanted to see if it could run in a browser with no server hop, so I exported the predictor to ONNX and ran it via onnxruntime-web with the WebGPU EP.What works: drop in an image, get a .ply you can download or preview live, all on your machine — your image never leaves the tab. The model is large (~2.4 GB sidecar) so first load is slow on a cold cache, but inference itself is a few seconds on a recent Mac.Caveats: SHARP's released weights are research-use only (Apple's model license, not the code's). I host the exported ONNX on R2 so thedemo "just works", but you can also export your own from the upstream Apple repo and upload locally.Happy to talk about it in the comments :)

Hacker NewsToolDevSource
0Sign in to voteCopy link

FL score

65

out of 100

Verdict

VALIDATE

high confidence

Competition

No competitor data yet

Trend

No signal yet

A browser-based implementation of Apple's SHARP 3D Gaussian splatting model using ONNX runtime for privacy-preserving, client-side 3D reconstruction from a single image.

The pain

Existing 3D reconstruction pipelines like Apple's SHARP require heavy server-side computation and complex setups, limiting accessibility and raising privacy concerns.

The gap

No easy-to-use, client-side, browser-based tool exists that runs SHARP or similar advanced 3D reconstruction models without server dependency or data upload.

Build angle

Focus on optimizing model size/load times and UX to enable seamless, privacy-first 3D reconstruction in-browser, targeting developers, researchers, and privacy-conscious users.

Strengths

  • Privacy-first approach: all processing happens locally in the browser
  • No server infrastructure needed, reducing costs and complexity
  • Leverages cutting-edge Apple research and ONNX runtime for accessibility
  • Open and extensible: users can upload their own models

Risks

  • Large model size (~2.4GB) causes slow initial load and may deter users
  • Research-use-only license limits commercial applicability
  • Niche developer/researcher audience limits market size
  • Performance and UX may suffer on less powerful devices or browsers
  • Unclear monetization or product-market fit beyond demo/tool status

Questions about this idea?

FlyBot reads the scoring and gives you a second opinion on “Apple's SHARP running in the browser via ONNX runtime web”.

Open FlyBot