Luma Concepts
Camera direction as dropdowns instead of prompt gambling
- luma_concepts
- luma_concepts
The Luma Concepts node exists for one reason: to keep you from praying that "slow push-in" makes it into your generated video. It's a small helper node that hands a set of explicit camera concepts to the Luma Text to Video and Luma Image to Video nodes, so your shot direction is a dropdown choice instead of a prompt that the model may or may not honor.
It doesn't call any API and it generates nothing. All it does is collect up to four camera concepts and pass them along as a typed bundle. That's the whole job - but it's a genuinely useful one, because Luma's models genuinely respond to these presets.
The inputs that matter:
- concept1 through concept4 - four dropdowns, each with the same list of camera moves. The list is long and readable:
zoom_in,zoom_out,push_in,pull_out,pan_left,pan_right,tilt_up,tilt_down,roll_left,roll_right,orbit_left,orbit_right,crane_up,crane_down,pedestal_up,pedestal_down,truck_left,truck_right,overhead,low_angle,high_angle,eye_level,pov,selfie,handheld,static,aerial,aerial_drone,dolly_zoom,over_the_shoulder,tiny_planet,ground_level,elevator_doors,bolt_cam, andNoneif you don't want to fill a slot. - luma_concepts - optional. A
LUMA_CONCEPTSinput so you can chain another Concepts node into this one and merge the lists. Handy when you want to compose camera direction from multiple sources.
The output is luma_concepts (type LUMA_CONCEPTS), which plugs into the matching input on Luma Video or Luma Image to Video.
The honest guidance: don't overthink it. Picking three conflicting moves ("zoom_in" plus "orbit_left" plus "handheld") is a real thing people do, and it usually reads as mush. Pick one dominant move and one supporting move - "push_in" + "low_angle" is a classic - and leave the rest at None. Also worth knowing: the concepts are camera-only. Character action, mood, and setting still live in your prompt; this node just owns the lens.
There's no installation or model involved - it ships with ComfyUI core. It's free to run, the cost only shows up when the video node downstream bills your Comfy account credits. And if you ever want the whole partner-node family out of your search results, --disable-api-nodes in the launch args does it.
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| concept1 | COMBO | 35 options: None, truck_left, pan_right, pedestal_down, low_angle, pedestal_up, +29 | |
| concept2 | COMBO | 35 options: None, truck_left, pan_right, pedestal_down, low_angle, pedestal_up, +29 | |
| concept3 | COMBO | 35 options: None, truck_left, pan_right, pedestal_down, low_angle, pedestal_up, +29 | |
| concept4 | COMBO | 35 options: None, truck_left, pan_right, pedestal_down, low_angle, pedestal_up, +29 | |
| luma_conceptsopt | LUMA_CONCEPTS | Optional Camera Concepts to add to the ones chosen here. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| luma_concepts | LUMA_CONCEPTS | — |