FLUX.2 [pro]

FLUX.2 [pro] is the model at the top of Black Forest Labs' current line, and the node's widget list is the tell: it's the most stripped-down image node in this pack. No steps, no scheduler, no CFG, no upsampling toggle - just prompt, dimensions, seed, and count. The pro tier is where BFL decided the model is good enough that your sampling knobs would only hurt, so the node gives you the settings that matter and keeps the rest server-side.

That's the right call, by the way. The pro tier is the hosted quality ceiling of the FLUX.2 family - the same line that produces the reference-based editing the open ecosystem was so impressed by - and most people were never going to improve on BFL's tuned defaults. You reach for this node when quality is the requirement and the cost per generation is acceptable. It's the "hero shot" node of the pack.

What you set

Required: positivePrompt, width, height (256–2048, stepped by 16). Then:

• referenceImages - the reference-based editing socket. This is where FLUX.2 earns its reputation: single-reference edits (restyle this subject, change this scene) work remarkably well from just one image.
• providerSettings.bfl.safetyTolerance - 0–5 moderation tolerance, default 2. It's the one knob besides prompt/dimensions that genuinely matters, because BFL moderation is strict and this is the hosted-pro escape hatch for borderline-but-legit content.
• seed - reproducibility (default 0 = random).
• numberResults - up to 20 variations.

That's the whole list. Output is one image, wired into Preview Image or Save Image like any other generator.

Pro vs the rest of the family

The FLUX.2 line is a ladder, and pro is the top rung: [pro] is the highest quality, [flex] is the cheap-and-fast tier, and klein is the open-weights family (with the non-commercial license on the 9B). If you're comparing them from ComfyUI, the question is cost per shot and what you need it for. Iterating → flex. Editing or quick drafts → klein. A deliverable where quality is the point → pro. The node even skips promptUpsampling, which the flex and Kontext-pro nodes have - at this tier BFL just doesn't think your prompt needs the help, and mostly it doesn't.

Install

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/Runware/ComfyUI-Runware
pip install -r ComfyUI-Runware/requirements.txt

Restart, or install "Runware" from ComfyUI Manager, and set your Runware API key (Settings → Runware API key or RUNWARE_API_KEY). No weights download - pro is a hosted product by definition, which is also why the licensing worries that follow the open weights never touch this node. The one real gotcha is cost: pro is the most expensive tier in the line and the node prints the bill on its title bar, so keep numberResults honest while you're iterating.