Attribution By Image Node

This is the oddest node in the BRIA AI API nodes pack, and possibly the most interesting one. Attribution By Image takes the images your workflow generated, shares them with BRIA over their API, and BRIA uses them to work out which data owners contributed to the generation - then pays those owners on your behalf. The images are erased from BRIA's side immediately after the calculation. It's a small piece of machinery that only makes sense once you know BRIA's business model: they train exclusively on licensed data, and this is the machine that closes the loop on paying the people behind that data.

You don't get a better image out of it. There's no IMAGE output - just an api_response string confirming what the API did. The point is legal and ethical plumbing, not pixels.

Should you use it?

Here's the important nuance from the pack's README, quoted nearly verbatim: this node is for workflows that use Bria's models - it's not necessary for Bria's API nodes. If you're running the rest of this pack (which all calls BRIA's hosted API), the API is already handling attribution on its end. The Attribution node exists for people who use BRIA's open-weights models locally, or otherwise generate images that should be attributed back to BRIA's licensed data owners.

So the honest advice: if you only use this pack's nodes, you probably don't need it - but it's a one-node addition that costs a trivial API call and makes the "who paid for the training data" question somebody else's problem. For a commercial product built on BRIA models, that's not nothing.

Inputs and output

• images - the generated image(s) to report. It batches, so you can wire in a whole folder of results.
• model_version - which BRIA model family produced them: 2.3, 3.0, or 3.2. Pick the one that matches what generated the images.
• api_key - your token (the BRIA_API_TOKEN placeholder gets rejected).

Output is a single STRING, api_response, containing whatever the attribution API returned (joined with --- for batches). Wire it into a Show Text / display node if you want to see it; otherwise it's a terminal node for a side branch.

Install

Standard for this pack:

cd path_to_comfyui/custom_nodes
git clone https://github.com/Bria-AI/ComfyUI-BRIA-API.git

or ComfyUI Manager → search "BRIA API" → restart. No models, no extra dependencies.

Gotchas

The README's API link points to BRIA's attribution documentation, and this endpoint does real work with real images - worth reading their attribution docs before you run it at scale, because "erased immediately after" is a claim you should confirm on their side if compliance matters to you. And one workflow trap: the model_version default is 2.3, so if your images came from a newer pipeline, bump it or the attribution math runs against the wrong model family.