Apple's SHARP running in the browser via ONNX runtime web
Details
- External ID
- 47995037
- Source
- HN
- Company
- —
- Product
- Apple's SHARP running in the browser via ONNX runtime web
- Website domain
- github.com
- Launched
- May 3, 2026
- Cohort
- —
- Upvotes
- 185
- Upvotes percentile
- 0.9563812600969306
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:26 p.m.
- Updated at
- Sept. 7, 2026, 9:26 p.m.
Description
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 :)
Enrichment
- Theme
- lightweight and on-device AI runtimes
- Vertical
- Horizontal
- Function
- Model & infra
- Audience
- Developer
- AI stance
- Not AI
- Project type
- Hobby / open-source project
- Normalized one-liner
- sharpe model runner for web
- Manually corrected
- False
Could you build this?
Partial While exporting a model and wiring ONNX Runtime Web to a WebGL/WebGPU Gaussian splat viewer is within reach of AI assistance, handling custom PyTorch operators, model quantization, and in-browser shader pipeline optimization requires deep graphics and ML systems knowledge.
What it would actually take: The stack relies on PyTorch for model export, ONNX Runtime Web (WebAssembly/WebGPU), and a Three.js/WebGPU-based 3D Gaussian Splatting rasterizer. The hardest parts are resolving unsupported PyTorch ops during ONNX export, tuning memory consumption to stay within browser limits, and achieving 60fps splat rendering on client hardware. This demands expertise in computer graphics shaders and ML model optimization.
Discussion
20 comments analyzed.
Competitors mentioned: 3D Gaussian Splatting (Kerbl et al., 2023), Polycam, Luma, Postshot, WorldLabs Marble model, Tencent Hunyuan HyWorld
Concerns raised: Model requires 2.4GB in browser, causing crashes on lower VRAM machines, Needs at least 8GB RAM, may not be enough even then, Only generates small depth sample from single image with holes when camera moves slightly, Quantization effectiveness unknown, model quality degradation unclear, Inference takes several seconds even on recent machines
Feature requests: Compressed model version for low-RAM machines, Full 360-degree orbital camera control, Double-click drag to move camera in non-orbiting mode
Competitors
Other products that read as similar to this one — 59 launches clear the similarity bar, closest 8 shown.
Attention rank: #8 of 60 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 181 days after the earliest competitor.
- Run TRELLIS.2 Image-to-3D generation natively on Apple Silicon · hn · 2026-04-20 · 202 upvotes · similarity 0.46
- I run AI background removal in the browser–no upload,no server · hn · 2026-04-10 · 7 upvotes · similarity 0.42
- Tsplat · hn · 2026-04-14 · 7 upvotes · similarity 0.40
- I made a Gemma 4 Mac app that names screenshots with local AI · hn · 2026-05-31 · 7 upvotes · similarity 0.39
- Sharpshot · ph · 2026-09-29 · 1 upvotes · similarity 0.39
- RunMat · hn · 2025-12-02 · 21 upvotes · similarity 0.39
- Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac · hn · 2026-07-29 · 919 upvotes · similarity 0.38
- Kiln · hn · 2026-07-28 · 9 upvotes · similarity 0.38
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a model & infra tool for Fintech yet.