PVL LumaPhoton Reframe (fal.ai)

You shot a portrait and the client wants a 16:9 banner. You have a 4:5 product shot and need a 9:16 reel. Cropping loses the subject; upscaling-and-cropping looks worse. Luma Photon's "reframe" is the model built for exactly this: it takes your image, fits it into a target aspect ratio, and paints in the new space so the result looks composed rather than cropped. This node wraps fal's luma-photon/reframe endpoint - image in, aspect ratio chosen, reframed image out, no local model.

It's part of the ComfyUI Assistant Node pack from pvlprk. The mechanism is the pack's standard fal flow: your image is base64-encoded and sent to the endpoint, the job is polled, and the result comes back as a normal IMAGE tensor.

How it works

The only required inputs are image and aspect_ratio (1:1, 16:9, 9:16, 4:3, 3:4, 21:9, or 9:21; default 16:9). The model repositions your content into that shape and regenerates whatever the new frame needs. Beyond that, the node exposes the endpoint's finer controls as optional inputs, all defaulting to -1 (meaning "not set" and not sent):

• prompt - an optional instruction for what to paint into the extended area.
• grid_position_x / grid_position_y - position the source image within an output grid.
• x_start / x_end / y_start / y_end - define a crop region of the source to keep and reframe, rather than the whole image.

The -1 default is a nice touch: leave them alone and the endpoint picks the framing; set them and you're steering exactly which part of the photo matters. Luma Photon is Luma AI's fast, efficient model, so these jobs are quick - this is the endpoint you'd reach for when you need a rapid reshoot-style reframe rather than a heavyweight edit.

The inputs that matter

• image - your source.
• aspect_ratio - the target shape; this is the whole job.
• prompt (optional) - describe what should fill the new space.

Output: one IMAGE, already at the target aspect ratio.

Installing it

Part of pvlprk/comfyui-pvl-api-nodes. Install via ComfyUI Manager (search "ComfyUI Assistant Node") or:

cd ComfyUI/custom_nodes
git clone https://github.com/pvlprk/comfyui-pvl-api-nodes

Set your fal.ai key as FAL_KEY (or FAL_API_KEY / FAL_CLIENT_KEY), restart, and make sure you have credits.

Where people get burned

The usual pack-wide traps - unset FAL_KEY, blank-placeholder errors hiding in the console. The reframe-specific gotcha: the model is reconstructing the scene, so content that's only partially in frame can get re-interpreted in the extended area - a half-visible arm might come back as a full arm you didn't have. If the result's composition isn't what you want, that's what the grid-position and crop-region inputs are for; a tight crop with x_start/x_end/y_start/y_end gives the model a lot less to invent.