TypeScript icon, indicating that this package has built-in type declarations

    0.2.1 • Public • Published


    This JavaScript implementation of Magenta's image models uses TensorFlow.js for GPU-accelerated inference.

    Complete documentation is available at https://tensorflow.github.io/magenta-js/image.


    Example Applications

    You can try our hosted demos for each model and have a look at the demo code.

    Supported Models

    Fast Arbitrary Image Stylization

    Implements Ghiasi et al.'s fast arbitrary style transfer model (paper, code). Wraps around Reiichiro Nakano's TensorFlow.js port of the model checkpoint.

    Getting started

    There are two main ways to get MagentaImage.js in your JavaScript project: via script tags or by installing it from NPM and using a build tool like yarn.

    via Script Tag

    Add the following code to an HTML file, and place a content (content.jpg) and style (style.jpg) image in the same directory:

        <!-- Load @magenta/image -->
        <script src="https://cdn.jsdelivr.net/npm/@magenta/image@^0.1.2"></script> 
        <img id="content" height="256" src="content.jpg"/>
        <img id="style" height="256" src="style.jpg"/>
        <canvas id="stylized" height="256"></canvas>
            const model = new mi.ArbitraryStyleTransferNetwork();
            const contentImg = document.getElementById('content');
            const styleImg = document.getElementById('style');
            const stylizedCanvas = document.getElementById('stylized');
            function stylize() {
              model.stylize(contentImg, styleImg).then((imageData) => {
                stylizedCanvas.getContext('2d').putImageData(imageData, 0, 0);

    Launch a simple HTTP server (e.g. python3 -m http.server) and point your browser to You should see your content and style images displayed and, after a few seconds, the stylized output.

    via NPM

    Add MagentaImage.js to your project using yarn or npm. For example, with yarn you can simply call yarn add @magenta/image.

    Then, you can use the library in your own code as in the following example:

    import * as mi from '@magenta/image';
    const model = new mi.ArbitraryStyleTransferNetwork();
    const contentImg = document.getElementById('content') as HTMLImageElement;
    const styleImg = document.getElementById('style') as HTMLImageElement;
    const stylizedCanvas = document.getElementById('stylized') as HTMLCanvasElement;
    function stylize() {
      model.stylize(contentImg, styleImg).then((imageData) => {
        stylizedCanvas.getContext('2d').putImageData(imageData, 0, 0);

    See style-transfer.glitch.me and our demos for example usage.

    Example Commands

    yarn install to install dependencies.

    yarn test to run tests.

    yarn bundle to produce a bundled version in dist/.

    yarn run-demos to build and run the demo.




    npm i @magenta/image

    DownloadsWeekly Downloads






    Unpacked Size

    2.15 MB

    Total Files


    Last publish


    • adarob
    • cghawthorne
    • hardmaru
    • iansimon
    • notwaldorf
    • vdumoulin