--- title: API | Luma Agents description: 'Call Luma''s image and video models directly: one request per generation, submit, poll, download.' --- The Luma API gives you direct access to Luma’s image and video models: [uni-1 and uni-1-max](/guides/model/index.md) for images, and [Ray 3.2](/guides/model/index.md) for video. Generate from a prompt or from reference images and keyframes, edit or restyle an existing image or clip, extend a shot, reframe a video to a new aspect ratio, or split an image into layers. Every job runs on the same async REST endpoint: you submit it, poll it, and download the output. 1. 1**Submit**`POST /v1/generations`Send the prompt and options 2. 2**Poll**`GET /v1/generations/{id}`Repeat until the state is completed or failed 3. 3**Download**`output[].url`A presigned link per output, valid for one hour | | Images | Video | | ----------------- | ----------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------- | | Make new | [From a prompt or reference images](/guides/images/generation/index.md) | [From a prompt, start and end frames, or keyframes](/guides/videos/generation/index.md) | | Change existing | [Edit or restyle](/guides/images/editing/index.md) | [Edit](/guides/videos/editing/index.md), [extend](/guides/videos/generation/index.md), [reframe](/guides/videos/reframing/index.md) | | Split into layers | [Layering](/guides/images/layering/index.md) | | You can also [upload a file once](/guides/files/index.md) and reuse it across many generations, and [delete a generation](/guides/generation-deletion/index.md) and its outputs when you no longer need them. If you’d rather have an agent pick the model, write the prompt and check the result, use [Agents](/agents/index.md).