CVAT vs Label Studio
Your data is what decides CVAT vs Label Studio. Images, video and 3D point clouds are CVAT's focus; Label Studio handles images, text, audio, video and time series, as well as model feedback work. Each one is free and open source.
| Feature | CVAT | Label Studio |
|---|---|---|
| License | MIT | Apache-2.0 |
| Paid cloud | A Team plan costs $23 to $33 monthly for each user | The Starter Cloud plan is $99 monthly, plus $49 for each extra user |
| Free option | Free plan and free self-hosting | Free open-source Community Edition |
| Data types | Images, video, 3D point clouds, audio | Images, audio, text, HTML, time series, video |
| AI help | SAM 2, SAM 3 and your own models | Your own model for pre-labeling; more in Enterprise |
| Quality checks | Ground truth, honeypots and consensus | Reviews and agreement in Enterprise |
| GitHub stars | About 16.9 thousand | About 28.5 thousand |
Choose CVAT when
- boxes, polygons, tracks or 3D cuboids make up most of your work
- you want ground truth and honeypots already built in for quality checks
- you need lots of export formats, for example COCO, YOLO and Pascal VOC
Choose Label Studio when
- you label images and also text, audio or time series
- you gather human rankings or corrections for model answers
- you want to plug your own model in for pre-labeling

Frequently asked questions on CVAT and Label Studio
Can you use CVAT for free?
Yes. Under the MIT License, CVAT Community costs nothing and is open source. The cloud version offers a free plan, and its Team plan starts at $23 for each user a month on a yearly plan.
What is the price of Label Studio?
The Community Edition costs nothing and is open source. Starter Cloud is $99 monthly, with $49 added monthly for every extra user.
Can I install CVAT or Label Studio on servers I own?
Yes, both. CVAT Community is a free edition you host yourself, and CVAT Enterprise works on-premises or fully air-gapped too. You can install Label Studio locally via pip or Docker, and the Enterprise edition is available in the cloud or on-premises.
Which tool works with text and audio?
Label Studio. Video, images, audio, text, HTML and time series can all be labeled, and human rankings of model answers are collected too. Images, video and 3D point clouds are CVAT's main focus, though it handles audio too.
Do Label Studio and CVAT sell labeling work as well?
Yes. CVAT also offers labeling done for you, with a team of more than 300 annotators. The company behind Label Studio, HumanSignal, offers data services from domain experts.
Which one offers more export formats?
More than 20 formats are supported for export by CVAT, including COCO, YOLO and Pascal VOC, so it works easily with most computer vision training code.
Send one brief, hear from three vendors
Describe your data once and we pass it to three annotation vendors that label that kind of data, so you can compare quotes side by side.
- One short form instead of three sales calls
- Matched on data type, volume and security needs
- You choose who to talk to