Luma Genie: 3D generator — what it does, formats, and rights to the result

Luma Genie is a cloud service from Luma AI built around the concept of AI agents. Its core task is generating and processing video, images, audio, and text. The developer describes it as a tool that lets agents explore, generate, and refine content across these four media while preserving brand style.

What it does

The service accepts text, images, video, audio, as well as style references and character references as input. This allows setting up fairly complex generation scenarios based on different types of source data.

The output is a video clip. According to the developer, the clip duration is 5 seconds. Available video resolutions: 540p, 720p, 1080, and 4k.

For generation, you can choose one of two model versions: Ray 2 (designated as ray-2) or Ray 2 Flash (ray-flash-2).

As for exporting static images, the authors show saving a file in JPG format in their code examples. No information about other export formats is provided in the fact sheet.

How to use it

Luma Genie is available as a web app and via API. The service does not distribute its model or weights, and it does not run locally — all generation happens on the Luma AI cloud side.

For API usage, the developer suggests reviewing the documentation. The authors provide a link to it in the news note about the Ray 3.2 model and API release, but the specific documentation URL is not given in the fact sheet.

Rights to the output

The provided fact sheet does not describe licensing terms or rights to generated content. This should be clarified directly in the service's user agreement.

Who it suits

The tool is aimed at those who need fast generation of short video clips with fine-tuning via style and character references. The mention of "ten directions before the morning call" and "fifty options by the end of the day" clearly points to intensive creative workflows — for example, in agencies or design teams where iteration speed matters.

The service is not suitable for those looking for a static 3D model generator in the conventional sense (mesh objects, OBJ or GLB formats). The output here is a video or a JPG frame, not a 3D scene. It is also not suitable for tasks requiring local deployment without internet access or passing your own model weights.

Luma Genie is a narrowly specialized video generator with an agent-based architecture, tuned for streaming visual content creation. It covers the need for fast video iterations but does not replace classic 3D modeling tools.