Runs in the browser · Vision

Run MODNet in your browser

MODNet downloads 11.3 MB over WebGPU (q4f16), or 6.3 MB over WebAssembly (uint8), to strip backgrounds from images entirely in the tab with Transformers.js. No install, no server.

Vision · Xenova/modnet

WebGPU download

11.3 MB

q4f16

WASM download

6.3 MB

uint8

Parameters: 6.5M
Pipeline task: background-removal

Reading

The WASM build downloads smaller here: 6.3 MB (uint8) against 11.3 MB (q4f16) for WebGPU. The CPU-fallback quant compresses tighter than the GPU pick for this model.

The smallest vision download in the catalog (11.3 MB).

The quant ladder spans 3.9x: 6.3 MB (uint8) to 24.7 MB (fp32).

Live run this browser

Checking what this browser can run…

All measured variants

Quant Download size

Variant	Size
uint8 WASM pick	6.3 MB
q8	6.3 MB
q4f16 WebGPU pick	11.3 MB
fp16	12.4 MB
bnb4	22.0 MB
q4	22.1 MB
fp32	24.7 MB

Sizes measured from the HuggingFace API file tree for Xenova/modnet, not estimated.

Use it with Transformers.js

import { pipeline } from "@huggingface/transformers";

const pipe = await pipeline("background-removal", "Xenova/modnet", {
  device: "webgpu", // usually falls back to "wasm"; wrap in try/catch for production
  dtype: "q4f16", // use "uint8" for the WASM build
});

const result = await pipe(/* your input */);

Requires npm install @huggingface/transformers (or the CDN build).

Embed this

A live badge for your model card or README: 11.3 MB over WebGPU, updated as the data is.

Markdown

$ [![MODNet in browser](https://localmodel.run/badge/browser/modnet/size.svg)](https://localmodel.run/browser/modnet)

HTML

<a href="https://localmodel.run/browser/modnet"><img src="https://localmodel.run/badge/browser/modnet/size.svg" alt="MODNet in browser" /></a>

Sources

Same job, different size

DINOv3 ViT-S/16 → 14.1 MB

FAQ

WebGPU or WASM for MODNet, in practice?

Chrome, Edge, and Safari 26+ run the WebGPU build (q4f16) on the GPU. Firefox has WebGPU on Windows and Apple Silicon Macs but not everywhere yet. Any browser without a WebGPU adapter falls back automatically to the WebAssembly build (uint8) on CPU: same model, slower to load, slower to run.

What does q4f16 mean for MODNet?

The WebGPU build here uses q4f16: 4-bit weights with some layers kept at 16-bit for stability: WebGPU's usual smallest clean build. The WASM fallback uses uint8: 8-bit integers: about a quarter the size of fp32, with a smaller quality trade than 4-bit.

Sizes measured 2026-08-01 from the HuggingFace API. Last validated 2026-09-14. See all browser models or the methodology.