在 OpenClaw 中使用 Luma。
Luma Ray 3.2 通过提示词驱动的变换修改现有视频，同时保持空间一致性和 3D 场景结构。发送源视频 URL 和描述变换的文本提示词——Luma 会以富有氛围的光照、环境细节以及物理一致的镜头运动重新渲染素材。OpenClaw agent 使用与聊天相同的 RunAPI 密钥调用它。
使用 RunAPI 通过 Luma Ray 3.2 修改视频。
要求：
- 从 RUNAPI_API_KEY 读取 API 密钥。
- 调用 POST https://runapi.ai/api/v1/luma/modify_video
- 将 model 设为 "luma-modify-video"。
- 将 source_video_url 设为要变换的视频的 URL。
- 将 prompt 设为描述所需变换的内容。
- 该任务为异步任务。轮询返回的 task_id 直到 status 为 "completed"。
- 完成后，从响应 output 中读取修改后的视频 URL。
curl -X POST https://runapi.ai/api/v1/luma/modify_video \
-H "Authorization: Bearer $RUNAPI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "luma-modify-video",
"prompt": "Transform the scene into a foggy autumn morning with golden light filtering through trees",
"source_video_url": "https://example.com/park-walk.mp4"
}'
{
"task_id": "tsk_abc123",
"status": "pending",
"model": "luma-modify-video"
}
三步在 OpenClaw 中使用 Luma
Configure RunAPI
Set RUNAPI_API_KEY in your environment. If you already configured RunAPI for chat in OpenClaw, the same key works for Luma video modification — no extra provider setup needed.
export RUNAPI_API_KEY=runapi_xxx
Call Luma modify_video
Send a POST request to the modify_video endpoint with model set to luma-modify-video, the source_video_url pointing to your input footage, and a prompt describing the transformation. Luma preserves the original scene geometry while applying the requested changes.
POST /api/v1/luma/modify_video
Poll for the result
The endpoint returns a task_id immediately. Poll the task status endpoint until the status is completed, then read the output video URL from the response. Video modification typically takes 30-120 seconds depending on complexity.
GET /api/v1/luma/modify_video/tsk_abc123
Luma modify_video API 参数
| 参数 | 类型 | 说明 |
|---|---|---|
model |
string |
Required. Use luma-modify-video. |
prompt |
string |
Required. Text description of the desired video transformation — style changes, lighting, atmosphere, or scene edits. |
source_video_url |
string |
Required. URL of the source video to modify. Must be a publicly accessible MP4 or MOV file. |
watermark |
boolean |
Optional. Whether to add a watermark to the output video. |
callback_url |
string |
Optional. Webhook URL that receives a POST when the task completes. |
OpenClaw 上的 Luma 是什么？
Luma Ray 3.2（Dream Machine）专注于以梦幻质感改造现有视频。它不从零开始生成，而是以提示词驱动的变换重新渲染源素材——大气调整、风格转换和环境改变——同时保持原有的空间布局和镜头运动不变。OpenClaw agent 通过 RunAPI 调用它，使用与聊天相同的 API key。
Luma 使用场景
氛围短片与音乐视频美学
用 Luma 的梦幻视觉质感将原始素材转化为风格化氛围短片。在保留原始运动的同时改变光效、氛围和色调，用于音乐视频概念或艺术项目。
广告方向的视觉快速假设验证
将现有产品视频以多个视觉方向重新渲染——电影感、极简风、黑色电影或季节性风格——在确定全面重拍前测试哪种风格最有效。
多工具流水线的基础素材
使用 Luma 对生成或拍摄的素材进行风格化处理，再传给其他工具做最终剪辑。Luma 的高速交付使其在更大型视频流水线中作为风格化步骤切实可行。
Luma + OpenClaw 常见问题
Luma modify_video re-renders existing footage with prompt-driven changes — atmospheric shifts (fog, rain, golden hour), style transfers (cinematic, anime, noir), lighting overhauls, and environmental transformations. It preserves the original camera motion and spatial layout of the scene.
Luma specializes in modifying existing video rather than generating from scratch. Its Ray 3.2 model maintains 3D spatial coherence during transformations, so objects keep their depth relationships and camera motion stays physically consistent. Models like Kling or Hailuo generate new video from text or images.
Asynchronous. The create endpoint returns a task_id immediately. Poll the task status endpoint or provide a callback_url to receive a webhook when the modified video is ready. Typical processing time is 30-120 seconds.
Luma uses flat per-task billing — each modify_video call costs a fixed amount regardless of video length. Check the RunAPI pricing page for current rates. Failed generations are not charged.
Yes. An OpenClaw agent can generate a video with Kling or Hailuo, then pass the output URL as source_video_url to Luma modify_video for post-processing — all through the same RUNAPI_API_KEY. Each step uses the same async polling pattern.