Superside Bria Background Standardizer (Hex Color)

Here's the thing nobody tells you before they hand you a background-swap node: the generative ones can't hit a specific color. You ask for a flat #F5F5F5 and they hand you an approximate grey with a gradient and a shadow, because re-lighting the scene is their whole trick. This node is the exception, and it's the one you want for anything that needs a consistent, exact background - eCommerce catalogues, avatar sets, brand-color shots, batch homogenization.

How it does it is worth understanding because it's what makes it deterministic. The node cuts the subject out using Bria RMBG 2.0 (fal's fal-ai/bria/background/remove endpoint - one API call, and it's a good cutout model, the kind of thing that handles hair decently). Then it composites the untouched subject onto a solid color locally, on your machine. No generative model ever touches the subject or the background pixels. No invented shadows, no lighting re-interpretation, no quality drift across a batch. The subject is bit-for-bit what went in; only the background pixels change.

Inputs are image, hex_color (a string like #F5F5F5, defaulting to #FFFFFF), and api_key. One optional worth knowing: edge_feather (0–15 pixels) softens the cutout edge so a jagged silhouette doesn't show against the flat color; 0 is a pixel-perfect hard edge. If your subject's hair or fur looks crunchy on the solid background, this is the knob, not a different model. sync_mode is the usual API plumbing. Outputs are image (the composited result, IMAGE) and info (STRING with the resolved hex and the source cutout URL).

A couple of notes for real use:

• The pack's README draws the line between the three Bria nodes explicitly, and it's the right mental model: exact solid color → this node. Generated scene → the two "Replace" nodes. The Replace nodes are generative and will subtly change the subject; this one can't and won't.
• Since compositing is local, the only per-image cost is the one background-removal call. That's the cheapest way to homogenize a batch - far cheaper than asking a generative model to reproduce a color.
• info gives you the cutout URL, so you can also think of this node as "background removal with a free solid-color composite."

It's still an API node, so the usual rules apply: fal credits, your image briefly leaves the machine for the cutout step, and blank api_key falls back to FAL_KEY (both blank = immediate error). There's no local background-removal fallback, so if you need fully offline cutouts, that's the KB's local stack - BiRefNet, rembg, InSPyReNet - instead.

Install - ComfyUI Manager (search "comfyui-superside-nodes") or:

cd ComfyUI/custom_nodes
git clone https://github.com/Superside/comfyui-superside-nodes
pip install -r requirements.txt

Restart ComfyUI, look under Superside. No model downloads - requirements.txt is fal-client, pillow, numpy, torch, requests. Community coverage of this pack is basically nil, so lean on the README, which is current and unusually specific.