FIBO - Generate Structured Prompt - Lite

This is the Lite twin of FIBO - Generate Structured Prompt, and the difference is all in the translator. Both nodes exist to give you the JSON that sits between your text and a FIBO image render - no pixels involved. The full version translates with Gemini 2.5 Flash; this one uses FIBO-VLM, BRIA's open-source bridge model, which is faster and cheaper but writes slightly less detailed structured prompts. If you're building pipelines on the structured-prompt format, Lite is the one you'll actually leave in your workflow, because the translation step is the cheapest part of the whole pipeline anyway - you don't need a frontier model to write decent JSON for a test run.

Like every node in the BRIA AI API nodes pack, it's a thin API client: your text goes up to BRIA's endpoint, a JSON string comes back. No image is generated, no render credits spent. You need a BRIA API token from platform.bria.ai.

How it works

prompt is POSTed to BRIA's structured-prompt endpoint (the Lite variant), and the response is the JSON that a FIBO generation node can consume. You can also feed images to get a structured prompt faithful to a reference image, or pass an existing structured_prompt plus new text to get an updated JSON - the same three input modes as the full node, just translated by the lighter bridge. Outputs: structured_prompt (STRING) and seed.

The output plugs straight into FIBO - Generate Image - Lite's structured_prompt input, so the natural loop is: Lite structured prompt → Lite generate → review → lock seed.

The inputs that matter

• api_token and prompt - required. The BRIA_API_TOKEN default is a placeholder that gets rejected; paste your real key.
• images - optional reference image(s).
• structured_prompt - optional existing JSON to update with new instructions.
• seed - string, default 123456.

Install

Same as the whole pack - there's no requirements file and no model download, just the clone:

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

or search "BRIA API" in ComfyUI Manager, restart, and set your token.

Lite vs. full, decided

The honest answer: use Lite unless you have a reason not to. The full node's Gemini translation shines on complex, multi-element scenes where you want the richest possible JSON. For anything you're iterating on - which is the whole point of looking at the JSON before rendering - the difference rarely matters, and Lite costs less. The one trap is identical across both: people assume the JSON is opaque and skip the whole "structured prompt" family, when actually it's the most transparent part of the pack. Wire it to a Show Text node once and you'll finally see what "structured prompt" means instead of taking the README's word for it.