Teachable Machine

Teachable Machine is a free web-based tool from Google that enables anyone to train custom machine learning models for image classification, audio recognition, and pose detection directly in a browser without writing code or requiring machine learning expertise, making artificial intelligence model creation accessible to students, educators, artists, and creative professionals. The image classification training accepts input through webcam capture, file uploads, or sample image collections, where users create categories by recording or uploading example images for each class they want the model to recognize, with the system training a neural network model in the browser using transfer learning on pretrained MobileNet architectures. The audio model training enables creating custom sound recognition models by recording audio samples for different categories, useful for detecting specific sounds, music genres, voice commands, or environmental audio for interactive applications and accessibility tools. The pose estimation training uses body pose detection to recognize different body positions and gestures, enabling gesture-controlled applications, fitness tracking, and interactive art installations that respond to human movement. The model export options include downloadable TensorFlow.js models for browser deployment, TensorFlow Lite models for mobile and edge devices, Python Keras models for server-side applications, and API endpoints for cloud-based inference, enabling trained models to power real applications across platforms. The real-time testing interface enables evaluating model accuracy before export by testing with live webcam, microphone, or uploaded test inputs. The training process takes seconds to minutes depending on data size, running entirely in the browser. Designed for educators, students, artists, makers, and anyone exploring machine learning without coding.