LabelmeChecked by Vetteroo review

Disclosure: Some links on this page are affiliate links. If you buy through them, we may earn a commission at no extra cost to you. We only recommend things we believe are useful. Read our full disclosure.

What it is

Desktop image annotation app with offline AI

Best for

People labeling images to train vision models

Price

$79 one-time (Pro)

Prices checked

October 31, 2026

We researched Labelme's official pricing, terms, product pages, help pages and outside reviews. We haven't used Labelme ourselves, so this is a research-based review, not a hands-on test.

Who it’s best for

• Machine-learning engineers labeling training data.
• Researchers who need offline labeling.
• Teams behind a firewall.

Who should skip it

• Probably not the best fit for teams that want cloud labeling with outsourced workers.

How it works day to day

This part draws on Labelme’s own product pages, help center and documentation, read October 31, 2026 (listed under Sources). We haven’t used Labelme ourselves.

Open a folder of images, click or draw a rough box, let the AI outline the shape, correct it, and move to the next image.

Key features in use

• SAM2 and SAM3 drafting
• YOLO-World
• Polygons and boxes
• JSON export
• Offline models

Setup and learning curve

Download the app for Mac, Windows or Linux; no Python needed.

Integrations

Works with AI agents such as Claude and ChatGPT; exports to YOLO and Pascal VOC with Pro.

Support

Email (support@labelme.io).

Real-world use cases

• Object detection datasets
• Segmentation masks
• Research projects
• Air-gapped labeling

What it costs

Prices checked October 31, 2026, from Labelme’s own pricing page unless noted. Prices change often, so check the official page before you buy.

• Pro: $79 one-time.
• Pro (Lifetime): $249 one-time.
• Team: $249 per seat (5+ seats you can reassign).
• 3-day free trial without a card; full refund within 14 days.

What we like

• One-time price.
• AI runs offline on your computer.
• No Python needed.
• From the author of open-source Labelme.

What to watch for

• Few outside reviews.
• Format conversion needs Pro and Python.
• Desktop only.
• No cloud collaboration.

The fine print

• Conversion to YOLO or Pascal VOC needs Pro and runs in a terminal with Python.
• If you connect an outside AI agent, that model may run in the cloud.
• Legitimacy signals we found: made by Kentaro Wada, author of the open-source Labelme project, which has about 16,000 GitHub stars; the desktop app is at version 7.9.0; its site shows ratings from 146 reviews.

The main alternative: Roboflow

Roboflow is a cloud platform for labeling images and training computer-vision models.

What it costs: Free plan; paid plans not checked

Pick Labelme if you…

• want offline labeling.
• want a one-time price.
• work with private images.

Pick Roboflow if you…

• want cloud collaboration.
• want model training built in.
• want a free plan.

Our take

Labelme is a trusted name in image annotation, and the paid app adds offline AI for a one-time price.

Side by side

From our company records. Check the official pages for current prices.

What we checked

• Labelme's pricing, product pages and terms, read October 31, 2026
• Trustpilot, searched October 31, 2026 (no usable score for Labelme)
• Roboflow's details: read on its official pages on October 31, 2026 (comparison only; not reviewed)
• The other official and outside pages listed under Sources (links checked October 31, 2026)

Sources

• Labelme (official site)
• Labelme pricing
• Labelme released versions
• Labelme open-source project (GitHub)

How we review

We read the company's own terms, test the sign-up or booking steps where we can, and compare it with alternatives. A commission never changes what we write. Read our disclosure.