Superside Bria Replace Background V2

This is the simpler member of the pack's Bria background family: Bria's "Replace Background V2" model on fal, driven by a text prompt and not much else. No reference image, no variations, no refinement toggle - just image + prompt + a few knobs, and a believable new background with realistic lighting and perspective comes back. If the fancier Superside Bria Background Replace node is the sports car, this is the sedan, and for a lot of jobs the sedan is the right call.

Inputs are image, prompt, and api_key as required, then negative_prompt (keep things you don't want out of the frame), steps_num (default 30, the sampling steps Bria runs), seed (-1 = random), and sync_mode. Outputs are image (IMAGE) and info (STRING, the fal result URL). That's the entire surface area. You can't show it a reference background, and you can't ask for four takes - you get one image per call and you re-run for variations.

That simplicity is its strength. It's one less thing to tune when you just need "put this portrait in a warm cafe" and you don't care about matching a specific reference look. But know what you're buying: like its bigger sibling, this is a generative background node. It re-lights the scene to sell the composite, which is what makes the result look real, but it can subtly alter the subject and it will absolutely not give you an exact flat hex background. The pack's README says it straight: prompt for #F2F2F1 and you'll get an approximate grey with a gradient or shadow. If your end goal is a catalogue-flat color, the Bria Background Standardizer is the node you actually want - it composites the cutout onto the exact color locally, no generative model in the path at all.

The trap to avoid: people grab whichever Bria node is on their canvas and wonder why a "background replace" came back with a grey-ish gradient instead of their brand color. That's not a bug, it's the difference between the generative family (this node and Bria Background Replace) and the standardizer. Decide before you wire anything: exact solid color = Standardizer, generated scene = this node.

Under the hood it's one fal call per image - your image uploads to fal's bria/replace-background endpoint, the model swaps and re-lights, the result lands back in the graph as a tensor. Metered fal credits, data leaves your machine, and an empty balance just stops the node. Nothing runs locally, so there are no weights to download.

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 and it's under Superside. Paste your fal key into the api_key widget; blank falls back to the FAL_KEY env var, both blank = immediate error. This pack has essentially no community tutorials - the README is thorough and current, so treat it as the manual.