Is Teachable Machine free?

Yes. You can gather samples, train, test, and export models without paying and without an account. A Google sign-in is optional and only needed to save a project to Google Drive. There are no paid tiers on the tool itself.

Do my images or audio get uploaded to a server?

No. Training and inference run on-device in the browser with TensorFlow.js, so webcam and microphone data stay on your computer. Data leaves only if you choose to save the project to Google Drive or use the quantized-model export converter.

What kinds of models can I train?

Three: image classification from files or a webcam, audio classification from short recorded sound samples, and pose classification from body positions. It is not built for object detection with bounding boxes or image segmentation.

Where can I use the exported model?

Exports are real TensorFlow.js models that run anywhere JavaScript does, including p5.js, ml5.js, Glitch, and Node.js. You can also export TensorFlow Lite builds and target hardware such as Coral and Arduino.

Is Teachable Machine suitable for production machine learning?

It is built for teaching and fast prototyping rather than large-scale production training. When you need bigger datasets, GPU training, or detection and segmentation models, compare the Replicate, AutoDL, and Coze overviews on Uwarp.

When should I skip Teachable Machine?

Skip it when the model needs bounding-box detection, segmentation, custom architectures, or training on datasets far larger than a few hundred samples, and when you need a managed endpoint serving many requests.

Is this page the official Teachable Machine site?

No. This is a curated overview with outbound links. Teachable Machine is a Google Creative Lab project; train models at teachablemachine.withgoogle.com.

Teachable Machine — train image, sound, and pose models in your browser

Capture image, sound, or pose samples and train a TensorFlow.js model in your browser—local-first, no sign-in.

Teachable Machine
No-Code ML
TensorFlow.js
Image / Sound / Pose

Publisher: Google Creative Lab
Type: No-Code Machine Learning
Pricing: Free
Reviewed: 13 September 2026

Official site

Quick verdict

Use when

Teach a classifier in minutes
Image, sound, or pose samples
Classroom and workshop demos
Prototype ML in the browser

Skip when

Large datasets and GPU training
Detection or segmentation output
Custom model architectures
Production serving at scale

Try instead

Teachable Machine vs common alternatives

Teachable Machine is a zero-setup browser trainer for image, sound, and pose classifiers; Roboflow targets computer-vision teams that need datasets and deployment, while Create ML builds native Apple models locally on a Mac.

Tap a dimension to focus

Pricing

Teachable MachineThis page
- Free, no account needed
- No paid tiers
Roboflow
- Freemium
- Paid for volume and deployment
Apple Create ML
- Free with Xcode
- Mac hardware required

Dimension	Teachable Machine	Roboflow	Apple Create ML
Pricing	Free, no account needed No paid tiers	Freemium Paid for volume and deployment	Free with Xcode Mac hardware required

Preview of Teachable Machine - not the live app. Confirm details on the official site.

Does this overview help you decide?

—

(—)

Learn more

Details below the decision summary—features, workflow, and scope notes.

What is Teachable Machine?

Capture image, sound, or pose samples and train a TensorFlow.js model in your browser—local-first, no sign-in.

What it costs

Tier snapshot before you start training on the official site.

Free tier: Yes
Pricing summary: Teachable Machine is free, and no account is required to gather samples, train a model, or test it in the browser. Signing in with a Google account is optional and is used to save a project to Google Drive; the quantized-model export converter is also the only flow that sends a few example images for processing. There are no paid tiers or credit meters on the tool itself. Confirm current hosting and export options on the official site—this page does not list currency amounts.

Reviewed on 13 September 2026 · Teachable Machine — official site

What Teachable Machine provides

Core capabilities of the public Google Creative Lab tool.

Three model types
Train image classifiers from files or a webcam, sound classifiers from short recorded clips, and pose classifiers from body positions.
Live capture or uploaded files
Gather samples by holding up objects to the camera, recording sounds, or striking poses, or import existing files for each class.
On-device training with TensorFlow.js
Training and inference run client-side, so you can work entirely on-device without webcam or microphone data leaving the computer.
Instant testing
After training, feed new examples to the model immediately to see how well it classifies before you export anything.
Export to web, mobile, and hardware
Download TensorFlow.js models for the web, TensorFlow Lite for apps and edge devices, or target boards such as Coral and Arduino.
Fits the JavaScript ecosystem
Exported models are real TensorFlow.js models, so they run with p5.js, ml5.js, Glitch, and Node.js as well as plain browser code.

How to use Teachable Machine

The Gather, Train, Export flow.

Gather samples
Open the site and start a new project, then create classes and record or upload examples for each one using the webcam, microphone, or files.
Train the model
Press train and wait for the browser to fit the model; open the advanced panel to adjust epochs, learning rate, and batch size when you need finer control.
Test with new examples
Point the preview at fresh input to confirm the classes separate correctly before relying on the result.
Export or host
Download the model for offline use, upload it to host online, or export a TensorFlow Lite or hardware target build for the device you are shipping to.
Save or share the project
Optionally sign in to save the project to Google Drive, or download the project file and share it with a class or teammate.

Who Teachable Machine is for

People who need a working model today, not a training pipeline.

Front-end and creative developers
Drop a classifier into a browser experience, prototype, or interactive installation without standing up training infrastructure.
Educators and workshop leads
Demonstrate how classification works end to end in one session, with an interface available in nineteen languages.
Product and design prototypers
Test whether camera or audio input is worth designing around before committing engineering time.
Makers and hardware tinkerers
Export models to Coral, Arduino, and TensorFlow Lite targets for physical sorting, control, and sensing projects.

When Teachable Machine is the right pick

Choose Teachable Machine when you want to show or test an idea quickly: capture a few classes, watch a model converge in the browser, and export it into a web page or a maker board the same afternoon. Skip it when the job needs large datasets, GPU training, bounding-box detection, or a model you have to serve at production scale.

Licensing and platform notes

Facts to verify on the official site before you depend on a workflow.

Browser-only, TensorFlow.js: Teachable Machine runs in a web browser on top of TensorFlow.js; there is no desktop install and no Python environment to configure.
On-device by default: Samples are not uploaded while you train or run inference. Data leaves the machine only if you save the project to Google Drive or run the quantized-model converter.
Google Creative Lab project: The tool is published by Google Creative Lab with an open community repository covering export snippets for TensorFlow.js, TensorFlow Lite, Java, and Python.
Model scope: Teachable Machine produces image, audio, and pose classifiers. It is not a bounding-box detector, a segmentation tool, or a general model-training service.

Frequently Asked Questions