Label Studio

Label Studio is a platform for teams preparing training datasets or reviewing AI outputs. It brings human annotation and evaluation into a browser interface that can run locally or on your own server.

• Data annotation: Label text, images, audio, video and time series, including object detection, segmentation, document OCR, transcription and speaker diarization.
• LLM and agent evaluation: Compare answers side by side, apply custom rubrics, collect preferences and corrections, and connect observability tools for human review of agent traces.
• RAG review: Assess retrieval relevance and grade generated answers against their source material.
• Custom review interfaces: Adapt layouts and templates to your data and evaluation criteria, including tasks that combine several data types.
• Model-assisted labeling: Connect machine learning models to supply preliminary labels and compare predictions, with APIs, a Python SDK and webhooks linking review work to training and evaluation pipelines.

Teams can keep datasets in separate projects and associate annotations with user accounts. Label Studio imports local files or data from Amazon S3 and Google Cloud Storage, and exports annotations for model training.

The open-source edition uses the Apache 2.0 license and supports local or server deployment, including Docker. A separate hosted cloud edition runs on the provider's infrastructure. Connecting cloud storage or external model services brings those providers into the data workflow.