Teachable Machine
Teachable Machine is an easy-to-use web-based tool for creating machine learning models with no expertise or coding required. It can be use...
Last verified:
What is Teachable Machine?
Teachable Machine is a free, web-based tool by Google that lets anyone create machine learning models quickly without writing any code. You can train a computer to recognize your own images, sounds, and poses using just your webcam or microphone, all live in your browser. The tool makes machine learning fast, easy, and accessible for everyone, from educators and students to artists and innovators.
The platform supports three project types: Image Project for classifying images, Audio Project for recognizing sounds and voice commands, and Pose Project for detecting body positions and gestures. After training, you can test your model in real-time through a preview pane, then export it as TensorFlow.js for web apps, TensorFlow Lite for mobile apps, or download it for local use. Projects can also be saved to Google Drive to resume later.
Teachable Machine is designed for educators, students, artists, makers, and anyone curious about AI with no prerequisite machine learning knowledge. All training happens locally on your device without sending images to Google servers, ensuring privacy. The tool is perfect for educational purposes, accessibility projects, art installations, gesture-based controls, voice-activated commands, and game development.
Teachable Machine pricing
Pricing model: Free
Teachable Machine is completely free to use with no paid plans or subscription tiers. There is no cost for creating projects, training models, exporting models, or hosting models online. All features including Image Project, Audio Project, Pose Project, Google Drive saving, and model export options are available at no charge.
Teachable Machine pros
- Completely free to use with no paid plans required
- No coding or machine learning knowledge required
- Train models entirely in-browser with one click
- Three project types: Image, Audio, and Pose
- Real-time testing through preview pane immediately after training
- All training happens locally on your device for privacy
- No images or data sent to Google servers
- Export models to TensorFlow.js, TensorFlow Lite, or download directly
- Save projects to Google Drive to resume later
- Works on desktop browsers like Chrome and Safari
- Quick training process takes only seconds to minutes
- Intuitive interface with clear step-by-step workflow
- Supports uploading images from files or using webcam
- Generate shareable public links for exported models
- Open-source code available on GitHub for learning
Teachable Machine cons
- Not officially supported on mobile devices (experimental only)
- Requires desktop browser for best performance and tutorial
- Browser compatibility issues on older browsers
- No official API access for programmatic use
- Limited to 3 classes by default in basic interface
- No advanced hyperparameter tuning beyond basic epochs/batch size
- Model accuracy depends heavily on sample quality and quantity
- This is an experiment, not an official Google product with guaranteed long-term support
Frequently asked questions about Teachable Machine
Is Teachable Machine free?
Yes, Teachable Machine is completely free to use. There are no paid plans, subscription fees, or hidden costs. All features including creating projects, training models, exporting models, and hosting them online are available at no charge.
Do I need to install anything to use Teachable Machine?
No installation is needed. Teachable Machine is a web-based tool that runs entirely in your browser. You just visit the website and start creating projects immediately.
Are my images stored on Google servers?
No. All training happens locally on your device. You train a neural network locally without sending any images to a server, which is why the tool responds so quickly.
Can I use Teachable Machine on mobile?
The mobile version is highly experimental. If you are new to machine learning, it is recommended to visit the site on a desktop for a helpful first-time tutorial and better performance.
What project types does Teachable Machine support?
Teachable Machine supports three project types: Image Project for training models with image samples, Audio Project for training models with sounds and voice, and Pose Project for training models to detect body positions and gestures.
How many samples do I need per class?
You should capture at least 30 images per class, but using 30-100 varied samples per class works well. The more data you have with different angles, positions, and variations, the better your model will learn the differences.
Can I export my trained model for use in my own projects?
Yes. You can export your model as TensorFlow.js for web apps, TensorFlow Lite for mobile apps, or download it for local use. You can also generate a shareable public link for your model or save it to Google Drive.
Do I need machine learning knowledge to use Teachable Machine?
No prerequisite machine learning knowledge is required. Teachable Machine is designed for educators, artists, students, innovators, and makers of all kinds – really anyone who has an idea they want to explore.
Why isn't my machine working the way I want it to?
This is common when exploring machine learning. Your machine doesn't have an understanding of higher-level concepts like faces or objects – it learns through the examples you give it. Try resetting your classes and capturing more varied samples from different angles and positions.
Who made Teachable Machine?
Teachable Machine was created by Google Creative Lab as a collaborative effort with friends from Støj, Use All Five, and the PAIR teams at Google. The code is open-sourced on GitHub and the image classification is powered by TensorFlow.js, developed by Nikhil Thorat and Daniel Smilkov.