Bria Remove Image Background

Let's be straight about this one: background removal is the most commoditized operation in this entire ecosystem. ComfyUI ships a local BiRefNet background-removal node in core, rembg is free and runs on CPU, and every checkpoint on earth can do a passable cutout. So why does Bria Remove Image Background - a paid API call to Bria's RMBG 2.0 model through the Comfy proxy - exist, and why would you use it?

Because RMBG 2.0 is one of the models that set the quality bar, and the API version means zero setup and zero VRAM. The edge quality - hair strands, fur, semi-transparent fabric - is the whole game in this category, and RMBG is genuinely good at it. If you need a batch of clean cutouts and you'd rather not install, download weights, and tune a local pipeline, this node is one image input away from done. Every partner-node rule applies: image goes to Comfy's proxy at api.comfy.org, Bria's servers run the model, credits come off your Comfy account, nothing runs on your machine.

Inputs are minimal, which is fitting for the job. image is it, plus seed (which, per the tooltip, controls whether the node reruns - results are nondeterministic regardless) and moderation, a DynamicCombo where "true" expands into visual_input_moderation and visual_output_moderation toggles. The node uploads your image to Comfy's storage first, then posts a remove-background request and polls a status endpoint until the result is ready. Output is a single IMAGE - the cutout - with transparency where the background was. That's the thing to remember downstream: the result has an alpha channel, so wire it into anything that respects transparency, or composite it over a new background with the usual nodes.

Honest guidance on when this is the right call. The background-removal doc's rule is to pick by which failure you can tolerate: flyaway hair, semi-transparent materials, batch throughput. If your source is a clean studio shot, local BiRefNet will match or beat RMBG for free, and this node is a waste of credits. If your source is a nightmare - fine hair, glass, smoke, pets, crowds - RMBG's edge quality earns its keep, and per-image pricing means a handful of hard cases is cheap. For thousands of simple product shots, the API bill will dwarf the local cost, and that's the trap.

The one grounded gotcha: like every Bria node, moderation is real. If an image gets flagged you get an error, not a cutout, and you still might have consumed the attempt - so don't build a fully unattended batch pipeline without checking a sample first.