Luma Dream Machine (fal)
Luma Dream Machine video, with keyframes and loops
- image
- end_image
- STRING
Luma's Dream Machine was one of the early closed video models that made people sit up - smooth, fluid motion at a time when everything else looked like a slideshow. LumaDreamMachine runs it inside ComfyUI through fal: text-to-video or image-to-video, with start/end keyframes and a loop option. Nothing runs locally; it's a hosted API call in gokayfem's ComfyUI-fal-API pack, billed per clip.
Reach for it when you want that characteristically smooth Luma motion, or when you specifically need a seamless loop or a controlled first-to-last-frame transition.
How it works
You choose a mode, give it a prompt, and optionally supply a start image and an end image. fal runs the generation on Luma's side and returns a URL to the clip. Like the other video nodes here, the output is a string URL - save it before the link expires.
The inputs and outputs that matter
prompt- describe the scene and the motion.mode-text-to-videoorimage-to-video. In image-to-video, your input image becomes the starting frame.aspect_ratio- a generous set:16:9,9:16,4:3,3:4,21:9,9:21.image(optional) - the start frame for image-to-video.end_image(optional) - the end frame. Set bothimageandend_imageand Luma interpolates a path between them, the same keyframe trick that makes for deliberate transitions.loop(default off) - turn it on for a clip that returns cleanly to its start. This is Luma's calling card for social loops and backgrounds.
The single output is a STRING - the video URL.
How to install it
Via ComfyUI Manager: search ComfyUI-fal-API, install, restart. Manually:
cd ComfyUI/custom_nodes
git clone https://github.com/gokayfem/ComfyUI-fal-API.git
pip install -r ComfyUI-fal-API/requirements.txt
No models download. Put a fal API key in config.ini under [API] (copy config.ini.example) or export FAL_KEY=your_key, restart, and find it under FAL/VideoGeneration.
Common issues & troubleshooting
- Auth error.
FAL_KEYnot set. Fix and restart. - Image-to-video ignored my image. Make sure
modeis set toimage-to-videoand theimageinput is connected - the mode switch is what tells Luma to use the frame. - The loop isn't seamless. Loops work best when the motion is cyclical to begin with; a shot with a strong directional move won't close cleanly no matter what.
- The URL expired. fal result links don't last - route the
STRINGoutput into the pack's Save-Media-from-URL utility or download it right away. - A run fails. The pack raises fal's real error message - content policy, quota, or a bad input. Read it.
- Cost. Every clip is a paid API call. The result cache serves repeat identical calls free (
force_rerunbypasses); the spend guard caps a session. - Windows Portable. Missing
fal-client?python_embeded\python.exe -m pip install fal-client.
Luma isn't always the highest-fidelity option next to the newest Kling, but for smooth motion, clean loops, and keyframe control it's a pleasant, predictable node to keep on the shelf.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | — | |
| mode | COMBO | text-to-video | 2 options: text-to-video, image-to-video |
| aspect_ratio | COMBO | 16:9 | 6 options: 16:9, 9:16, 4:3, 3:4, 21:9, 9:21 |
| imageopt | IMAGE | — | |
| end_imageopt | IMAGE | — | |
| loopopt | BOOLEAN | false | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| STRING | STRING | — |