---

Installation

npm i nsfw-filter

Usage

1import NSFWFilter from 'nsfw-filter';
2
3const isSafe = await NSFWFilter.isSafe(image);

Full example

1import { useState } from 'react';
2import NSFWFilter from 'nsfw-filter';
3
4function ImageUploader() {
5  const [imageUrl, setImageUrl] = useState('');
6
7  const handleImageUpload = async (event) => {
8    const file = event.target.files[0];
9
10    // Check to see if the image is appropriate
11    const isSafe = await NSFWFilter.isSafe(file);
12    if (!isSafe) return 'Image is not appropriate';
13
14    // Process the image if it is safe
15    if (file) {
16      const reader = new FileReader();
17      reader.onloadend = () => {
18        setImageUrl(reader.result);
19      };
20      reader.readAsDataURL(file);
21    }
22  };
23
24  return (
25    <div>
26      <input type="file" onChange={handleImageUpload} />
27      {imageUrl && <img src={imageUrl} alt="Uploaded" />}
28    </div>
29  );
30}
31
32export default ImageUploader;

Real world usage

nsfw-filter is currently used in production to process hundreds of thousands of images for a popular image restoration service called restorePhotos. It helps prevent people from uploading inappropriate pictures. See how it's used here.

Using in a browser environment with vite

you can polyfill the core node modules used in nsfw-filter for browser compatibility such as

["path", "stream", "assert", "events", "zlib", "util", "buffer"]

more on this: https://vitejs.dev/guide/troubleshooting.html

install vite-plugin-node-polyfills with npm: https://www.npmjs.com/package/vite-plugin-node-polyfills in vite.config.js

1
2import { defineConfig } from "vite";
3import react from "@vitejs/plugin-react-swc";
4import { nodePolyfills } from "vite-plugin-node-polyfills";
5import path from "path";
6export default defineConfig({
7  plugins: [
8    nodePolyfills({
9      // To add only specific polyfills, add them here. If no option is passed, adds all polyfills
10      include: ["path", "stream", "assert", "events", "zlib", "util", "buffer"],
11    }),
12    react(),
13  ],
14  resolve: {
15    alias: {
16      "@": path.resolve(__dirname, "./src"),
17      "~": path.resolve(__dirname, "./"),
18    },
19  },
20});
21

bonus: you can polyfill path to support path aliases for vite when you deploy your could on the cloud for example vercel. so you can use @ import { cn } from "@/lib/utils" disclaimer: using nsfw-filter on the browser will dramatically increase your bundle size.

How it works

This library uses both Tensorflow.js, an OSS library for machine learning models, and nsfwjs to predict whether a given image is NSFW (Not Safe For Work).