Runs in the browser · Vision
Run BiRefNet Lite in your browser
BiRefNet Lite downloads 109.2 MB over WebGPU (fp16), or 109.2 MB over WebAssembly (fp16), to strip backgrounds from images entirely in the tab with Transformers.js. No install, no server.
Vision · onnx-community/BiRefNet_lite-ONNX
- Parameters
- 44.36M
- Pipeline task
- background-removal
Reading
Only fp16 and fp32 builds are published for this repo, no quantized variant yet.
2.46 MB/M params against a 1.73 MB/M vision median: dense for its parameter count.
Will it run in your browser?
Checking for WebGPU support…
All measured variants
| Variant | Size |
|---|---|
| fp16 WebGPU pick WASM pick | 109.2 MB |
| fp32 | 213.6 MB |
Sizes measured from the HuggingFace API file tree for onnx-community/BiRefNet_lite-ONNX, not estimated.
Use it with Transformers.js
import { pipeline } from "@huggingface/transformers";
const pipe = await pipeline("background-removal", "onnx-community/BiRefNet_lite-ONNX", {
device: "webgpu", // usually falls back to "wasm"; wrap in try/catch for production
dtype: "fp16", // use "fp16" for the WASM build
});
const result = await pipe(/* your input */);
Requires npm install @huggingface/transformers (or the CDN build).
Sources
Same job, different size
FAQ
WebGPU or WASM for BiRefNet Lite, in practice?
Chrome, Edge, and Safari 26+ run the WebGPU build (fp16) 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 (fp16) on CPU: same model, slower to load, slower to run.
What does fp16 mean for BiRefNet Lite?
The WebGPU build here uses fp16: 16-bit floating point (half precision): smaller than fp32 with effectively no quality loss. The WASM fallback uses fp16: 16-bit floating point (half precision): smaller than fp32 with effectively no quality loss.
Sizes measured 2026-08-01 from the HuggingFace API. Last validated 2026-08-03. See all browser models or the methodology.