Teachable Machine

Teachable Machine is a free, web based tool from Google for creating machine learning models without code. Users gather examples for each class they want recognised, train the model in the browser with one click, and test it live before exporting it for use in their own sites, apps and physical projects. The current version replaced an earlier 2017 edition, which remains available. It supports three kinds of model. Image models classify pictures from uploaded files or a webcam, audio models classify short recorded sound samples, and pose models classify body positions from files or live webcam poses. Training uses transfer learning on top of pretrained networks, with MobileNet behind the image and pose models and the Speech Command Recognizer behind the sound model. It suits educators, students, artists, makers and prototypers; Google links lessons on AI ethics and machine learning concepts, and showcases projects ranging from an Arduino sorting robot to an assistive communication setup triggered by facial gestures. Teachable Machine is free to use, and Google publishes no paid tier. Privacy is handled on device. Training runs in the browser tab without sending samples to any server, and nothing is saved unless the user chooses to download the model or save the project, including all samples, as a zip file in their own Google Drive. Models are TensorFlow.js models that can be hosted on Teachable Machine for use in any website or app, or converted to TensorFlow and TensorFlow Lite for local use; hosting publishes only the model, not the training examples. The tool runs on desktop Chrome or Safari and shows an unsupported message on other browsers and devices.