# Overworld

> Overworld is an open-science AI research lab building interactive world models — AI-generated video environments that respond in real time to text prompts and controller input at game-level framerates. Its open-source World Engine is a Python inference library that loads Waypoint world models and LoRA adapters from HuggingFace to generate playable, promptable worlds locally or in the browser.

This llms.txt was generated by the API Evangelist enrichment pipeline from the Overworld public developer surface. Overworld does not publish a REST API; its primary developer artifact is the open-source `world_engine` Python library.

## Product
- [Overworld](https://over.world/): Interactive world models — AI-generated, promptable, controllable virtual environments.
- [World Engine (GitHub)](https://github.com/Overworldai/world_engine): First-party Python inference library (WorldEngine + CtrlInput classes). Install: `pip install world_engine @ git+https://github.com/Overworldai/world_engine.git`
- [Overworld Stream (browser)](https://overworld.stream): Browser-based experience for interacting with generated worlds.

## Docs
- [World Engine Documentation](https://overworld-world-engine.readthedocs-hosted.com/en/latest/index.html): Quickstart, API reference (WorldEngine, CtrlInput), quantization, and modeling internals.
- [Install / Local Play](https://over.world/install): Local installation instructions.
- [About](https://over.world/about): Company, mission, and team.
- [Blog](https://over.world/blog): Research and product updates.

## Code
- [Overworld on GitHub](https://github.com/Overworldai): Organization with the World Engine and supporting open research repositories.
- [Biome](https://github.com/Overworldai/Biome): Local world client interface (TypeScript) for running Waypoint world models.

## Legal
- [Privacy Policy](https://over.world/privacy)