Infinigen Indoors is a Blender-based procedural generator for creating photorealistic indoor scenes. Building on the existing Infinigen system, which focuses on natural scenes, Infinigen Indoors introduces a diverse library of procedural indoor assets, including furniture, architecture elements, appliances, and other daily objects. It also includes a constraint-based arrangement system that allows users to specify scene composition constraints using a domain-specific language and a solver that generates scene compositions that maximize constraint satisfaction. The system provides an export tool for generating 3D objects and scenes that can be used in real-time simulators like Omniverse and Unreal for training embodied agents. Infinigen Indoors is open-sourced under the BSD license. The paper discusses the system's contributions, including the library of procedural generators, the constraint-based arrangement system, and the export tool, and presents experiments validating the effectiveness of the generated data in tasks such as shadow removal and occlusion boundary detection.Infinigen Indoors is a Blender-based procedural generator for creating photorealistic indoor scenes. Building on the existing Infinigen system, which focuses on natural scenes, Infinigen Indoors introduces a diverse library of procedural indoor assets, including furniture, architecture elements, appliances, and other daily objects. It also includes a constraint-based arrangement system that allows users to specify scene composition constraints using a domain-specific language and a solver that generates scene compositions that maximize constraint satisfaction. The system provides an export tool for generating 3D objects and scenes that can be used in real-time simulators like Omniverse and Unreal for training embodied agents. Infinigen Indoors is open-sourced under the BSD license. The paper discusses the system's contributions, including the library of procedural generators, the constraint-based arrangement system, and the export tool, and presents experiments validating the effectiveness of the generated data in tasks such as shadow removal and occlusion boundary detection.