Luma: Text to Video
High-quality videos can be generated using simple prompts.
简介
This ComfyUI workflow turns plain text prompts into short, high-quality videos using Luma's text-to-video model. The core of the graph is the LumaVideoNode, which sends your prompt and settings to Luma's API and polls for the rendered result. The LumaConceptsNode feeds structured "concepts" into LumaVideoNode, giving you fine control over camera behavior like pans, tilts, or zooms without overloading the prompt text. A SaveVideo node writes the returned video to disk, while a MarkdownNote provides in-graph guidance and tips.
你需要提供
- 正在读取工作流
在 ComfyUI 中运行。 此工作流无法在本页面运行。你可以在 Comfy Cloud 中打开（已预先加载所需内容），或下载后在自己的 ComfyUI 中使用。
在 Cloud 中试用常见问答
In this workflow, seed is only used to decide whether LumaVideoNode should re-run (invalidate cache). It does not control Luma’s internal randomness, so outputs can vary between runs even with the same prompt and concepts.
Use LumaConceptsNode to specify camera behavior (e.g., pan, tilt, zoom). Keep movement instructions in the concepts object and avoid duplicating or contradicting them in the text prompt. Start with a single movement at a modest intensity and iterate.
SaveVideo writes the file to the path you set in its parameters. The Luma API returns a ready-to-save video (commonly MP4), and SaveVideo stores that file without additional conversion.
Use the API in a secure environment: allow access from 127.0.0.1/localhost, call it from HTTPS contexts when embedding in web services, and verify you can reach the Luma API from your machine. Check firewall/proxy settings, confirm your API key, and re-run.