Superside Bria Background Replace

Bria's newer generative background-replace model, wrapped for ComfyUI. If you want to drop a product or portrait onto a different background - a studio set, an outdoor scene, a room - this node is the richer of the pack's two generative options: it can take a reference image as well as a text prompt, it refines your prompt for you, it has a fast/quality toggle, and it'll hand you up to four variations in one call.

The required inputs are the familiar image, prompt, and api_key. The optional list is where it gets interesting:

• ref_image - a reference background to guide the look, used instead of or alongside the text prompt. This is the node's standout feature over the older V2 node. Point it at "a bright airy kitchen" photo and Bria tries to match that vibe rather than whatever your words conjure.
• negative_prompt - what you don't want in the new background.
• num_images - 1 to 4 variations per run.
• refine_prompt - on by default; lets Bria expand/improve your prompt internally. Generally leave it on.
• fast - on by default; flip off for the higher-quality (slower) model when the subject is tricky.
• seed - -1 = random, set it to reproduce a take.

Outputs are image (IMAGE) and info (STRING, the result URL). Straight into a Save Image or straight into the next node - nothing exotic.

The important caveat, and the pack's README is admirably blunt about it: this is a generative background node. It re-lights the scene, which is exactly what makes it look real, but that means it may subtly alter the subject, and it cannot produce an exact flat hex background. If you prompt for #F2F2F1, you get an approximate grey with a gradient or a shadow, not the precise value. The pack's own decision table says it plainly: for an exact catalogue-flat color, use the Bria Background Standardizer instead; reach for this node when you want a scene, not a color. That distinction is the whole game when you're choosing between the three Bria nodes in this pack.

Mechanically it's a straightforward fal call: your image uploads, Bria segments and re-renders the background, and the result comes back as a tensor. One call per image, metered fal credits, your input leaves the machine. For batch work, num_images lets you generate alternatives in one shot and pick - cheaper than four separate runs of trial-and-error.

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, find it under Superside. Paste your fal key into the api_key widget (blank falls back to FAL_KEY, both blank errors immediately). No model downloads - the only new dependency in requirements.txt is fal-client. And because this pack has no community footprint to speak of, the README is the manual; it's current and unusually thorough.