BYOKey Bria Image

Bria is the less-hyped player in this pack, and that's fine - it's not aimed at the same crowd. It's an API-first image generation service pitched at commercial/enterprise use: licensed training data, predictable behavior, safe-by-default output for marketing and product pipelines. BYOKey_Bria_Image calls Bria's text-to-image endpoint with your key (sent as the api_token header) and returns a ComfyUI IMAGE. If you need a hosted image generator your organization can actually justify, this is the "boring and reliable" option on the roster.

The README has a genuinely interesting implementation note: Bria's official ComfyUI integration only ships edit-class nodes, not a text-to-image node. So this BYOKey node implements the public text-to-image endpoint that Bria itself doesn't expose through its own ComfyUI nodes. That's the whole value-add - it fills a gap rather than duplicating one.

How it works

It POSTs to {base_url}/text-to-image/{model} with your prompt, num_results, and optional seed (only sent if you set a positive value). Bria returns one or more image URLs, which the node downloads and stacks along the batch dimension into a single IMAGE tensor. So asking for 4 results gives you a batch of 4, exactly like a local node returning multiple images.

The inputs that matter

• prompt - the text description of the image. Everything about the result starts here; Bria's model is prompt-faithful but conservative.
• model - a free-text field defaulting to bria-2.3 (the tooltip example). Bria releases numbered model versions, and since this is a string you can point at a newer one when it ships.
• num_results - 1 to 4 images per call. Useful for picking the best of a small set.
• seed - optional; leave at 0 and it's unset.
• base_url - defaults to https://engine.prod.bria-api.com/v1; override only for relays.

Output is image (IMAGE), wired into a preview or the next node in your pipeline.

Installing it

It's one node in the comfyui-byokey pack, which installs with zero extra dependencies - just what ComfyUI bundles (torch, aiohttp, PIL, numpy, av). Use ComfyUI Manager (search "byokey") or:

cd ComfyUI/custom_nodes
git clone https://github.com/MeteorAndy/comfyui-byokey.git

Restart, and it's under api/byokey/image. No model downloads, no GPU requirement - Bria does the rendering.

Gotchas

The honest caveat is that Bria's image quality doesn't headline against the frontier - it's not trying to be the best renderer, it's trying to be the safest commercially-defensible one. If you want maximum quality per dollar you're better off with a different node in this pack; if you need consistent, licensable, brand-safe images for actual business use, this is the one. The other thing to know is Bria's content policy travels with the API - anything it refuses, the node refuses, and a third-party relay with "looser" moderation doesn't change what Bria's model does. Pack-standard hygiene: your key lives in the widget and is saved into the workflow JSON, so scrub it before sharing. If Bria is the API your work already pays for, this is the clean way to get it into ComfyUI.