Label Studio
Open-source data labeling for text, images, audio, and video
Label Studio is a flexible annotation platform. Teams use it to label data for machine learning, such as drawing boxes on images, tagging entities in text, transcribing audio, and rating language model outputs. Its configuration system lets users define custom labeling interfaces with simple markup, so one tool can serve a wide range of projects, and it runs locally or on a server.
Model-assisted labeling speeds up work. A machine learning backend can pre-label data for humans to correct, and active learning helps prioritize the most informative examples. The platform supports evaluating language models, with templates for tasks like ranking responses and checking for factual errors, which has become a common use as teams build AI applications. An API and SDK connect it with training pipelines.
The community edition is free and open source, and HumanSignal sells Label Studio Enterprise with features such as role-based access, workflows, analytics, and security. Self-hosting needs some technical work. Alternatives include CVAT for computer vision, Labelbox, Scale AI and Encord as commercial platforms, Prodigy for text, and Argilla for language model data. Label Studio suits teams that want a flexible, self-hosted annotation tool.