Prodigy

Prodigy is a downloadable, extensible annotation tool from Explosion for creating training and evaluation data for machine-learning and NLP systems. It runs on a user’s own hardware rather than as a hosted SaaS product, and the official site emphasizes that data does not leave the user’s machine. Its browser-based annotation app supports focused one-task-at-a-time review, custom recipes, active-learning workflows, information extraction, text classification, image and audio annotation, and integration with the spaCy ecosystem. Prodigy uses a lifetime license paid once, with individual and team options.

✨ Key Features

• Local-first annotation — Run the tool on your own machine or infrastructure.
• Text classification — Label documents and text categories.
• Information extraction — Annotate entities and relations for NLP.
• Named-entity recognition — Build and review entity data for models.
• Image annotation — Label visual data for computer-vision workflows.
• Audio annotation — Work with speech and audio datasets.
• Active learning — Prioritize examples that help improve a model.
• Custom recipes — Configure annotation interfaces for a project.
• Browser app — Let annotators review focused tasks in a web interface.
• Programmatic workflows — Import, export, and extend data through code.
• spaCy integration — Use Prodigy with Explosion’s NLP ecosystem.
• Privacy by architecture — Keep sensitive data on controlled hardware.

💳 Pricing

• Commercial lifetime license — A one-time purchase for the licensed version, with individual and team options; this is not a hosted freemium subscription.
• Academic exception — Degree-awarding academic institutions can request an interim free license for eligible academic research, subject to the vendor's restrictions. This is not a general trial for small projects.
• Updates and support — Confirm the update period, seats and support terms for the selected license.

⚖️ Alternatives Compared

🎯 Use Cases

• NLP dataset creation: Prepare text for classification and extraction models.
• Entity labeling: Build named-entity recognition training data.
• Computer vision: Annotate images for visual models.
• Speech datasets: Review and label audio examples.
• Active-learning loops: Improve model quality with targeted examples.
• Sensitive data projects: Keep annotation on private infrastructure.
• Research prototyping: Test labeling recipes quickly.
• spaCy projects: Create data for custom NLP pipelines.