Bria Tailored Gen

Every company with a product catalog eventually asks the same question: can we generate on-brand images at scale, without a photographer and without the model drifting into generic-AI look? BRIA's answer is Tailored Gen - a node that generates images from a tailored model, which is BRIA's term for a model trained on your specific visual IP: your product line, your character, your brand's illustration style. You give it the model ID, a prompt, and it reproduces your IP instead of approximating it.

This is a V1 (legacy) node in the BRIA AI API nodes pack, hitting BRIA's /v1/text-to-image/tailored/ endpoint. Like every node in the pack it runs on BRIA's cloud, so you need an API token from platform.bria.ai - and, unlike most of the pack, a trained tailored model ID to point it at, which you get by training one on BRIA's platform (there's a Tailored Model Info node in this pack that retrieves a model's defaults).

How it works

The request is a straightforward text-to-image call: your prompt gets prefixed with the model's generation_prefix (BRIA's way of forcing the model back into your IP's style), sent to the tailored endpoint, and the result comes back as an IMAGE tensor. Simple enough - the complexity is all in what you paid BRIA to train, not in this node.

The inputs that matter

• model_id and api_key - required. The model ID comes from BRIA's platform after you train your tailored model.
• generation_prefix - this is the secret sauce. It's the text BRIA prepends to every prompt to keep generations on-style; the Tailored Model Info node can fetch the right value for your model so you don't have to guess it.
• prompt - what you want in your style. Leave it empty and you still get pure-style generations.
• model_influence (default 1.0) - how hard the tailored model pushes against the base generator. Lower it if generations look too "on-model" and not enough like your text prompt.
• guidance_method_1 / guidance_method_2 - optional ControlNet-style guidance: controlnet_canny, controlnet_depth, controlnet_recoloring, or controlnet_color_grid, each with a scale and an image input. Handy for forcing composition (canny/depth) or color grading (recoloring/color_grid) while staying on-brand.
• steps_num (8), fast, seed (−1 = random), aspect_ratio (default 4:3) - the standard tuning dials.

Install

Same as the whole pack - Manager (search "BRIA API") or:

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

Restart, add your token, and you're generating - after you've trained a tailored model on BRIA's side, which is the actual prerequisite the README glosses over.

Reality check

Tailored Gen is where the "licensed data" pitch pays for itself or falls flat, depending on your IP: it's built for reproducible, legal, on-style commercial output, not for novelty. The gotchas are mostly upstream - the model has to be trained well, and generation_prefix and fast/steps_num are values the Tailored Model Info node exists to fetch because guessing them gives mediocre results. There's a tailored_workflow.json in the pack's workflows folder showing the intended wiring. If you're a hobbyist with no IP to protect, this node is a solution looking for a problem; if you're a brand manager, it's arguably the single most useful thing in the pack.