← Retour au blog
tech 3 May 2026

Running Apple's Sharp Model in the Browser via ONNX Runtime Web

Discover how the ml-sharp-web project leverages ONNX Runtime Web to run Apple's Sharp model directly in the browser, transforming how we interact with ML models.

Article inspired by the original source
Show HN: Apple's Sharp Running in the Browser via ONNX Runtime Web ↗ github.com

Introduction

Running machine learning models directly in the browser is a major breakthrough, offering unprecedented possibilities for developers and tech companies. The "ml-sharp-web" project highlights this innovation by enabling the Apple Sharp model to run via ONNX Runtime Web. But what does this really mean for you? Let's dive in!

Understanding the Apple Sharp Model

Apple Sharp is a machine learning model designed to create Gaussian splats, a technique often used in image processing to blur or soften edges. Traditionally, running such a model required significant server resources. However, the integration of ONNX Runtime Web changes the game.

ONNX Runtime Web: An Overview

ONNX (Open Neural Network Exchange) is an open format that facilitates interoperability between diverse ML frameworks. The ONNX Web runtime allows these models to run directly in the browser without server dependency. This means reduced infrastructure costs and improved latency.

Key Advantages

  1. Portability: Use the same model across multiple platforms.
  2. Cost Reduction: Less server dependency.
  3. Latency Improvement: Faster response times.

Use Cases: Why Does It Matter?

A concrete example is the e-commerce domain. Imagine a clothing retail site using the model to provide real-time previews of items worn by virtual mannequins. By running the model in the browser, the user experience is instant and uninterrupted.

How Does It Work?

Model Integration

The ml-sharp-web project uses ONNX to convert the Apple Sharp model into a browser-compatible format. Through JavaScript scripts, the model is loaded and executed directly client-side. This has been made possible thanks to the open-source community, notably the GitHub repository [bring-shrubbery/ml-sharp-web](https://github.com/bring-shrubbery/ml-sharp-web).

Challenges and Solutions

One of the main challenges lies in optimizing performance. Browsers are not initially designed for heavy ML workloads. However, techniques such as model size reduction and the use of WebAssembly (Wasm) help overcome these hurdles.

Conclusion

The integration of ONNX Runtime Web to run the Apple Sharp model directly in the browser opens new perspectives for developers and companies. It's an advancement that facilitates the creation of more responsive and cost-effective applications.

Let's discuss your project in 15 minutes.

ONNX Runtime Web Apple Sharp model machine learning browser execution Gaussian splats
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call