Zhenzhen_Flux_2_Pro

Flux 2 Pro is BFL's professional tier - the one aimed at "this needs to be right" work rather than "this needs to be amazing" work - and Comfly_Flux_2_Pro is the Comfly pack's paid client for it. Like its Max sibling, the node is deliberately bare: prompt in, image out, no sampling dials. Where Max leans into quality per generation, Pro is the steadier, more predictable tier for production runs. And like everything in this pack, the model runs on BFL hardware through the Comfly relay - you pay per image, your GPU watches.

Inputs that matter

• prompt - required, multiline. There's no prompt-upsampling here, so write the prompt you actually want generated; on a per-call paid node that discipline pays off.
• input_image … input_image_8 - up to eight reference images. This is the main differentiator available to you: drop in a reference and the prompt becomes an edit or reimagination; stack references to blend characters or styles. For a professional tier, this multi-image input is what makes it useful for real production assets.
• width / height - 64–6000, step 8, default 1024 square.
• safety_tolerance - 0–5, default 2, the API's content-filter dial. Leave it at 2 unless you have a specific reason not to.
• output_format - png (default) or jpeg. Pro defaults to png - sensible, since professional pipelines usually want lossless for downstream compositing.
• seed - -1 for random, or lock for reproducibility.

Outputs: image (IMAGE), image_url, and response.

Install

The usual Comfly drill:

cd ComfyUI/custom_nodes
git clone https://github.com/ainewsto/Comfyui_Comfly

Or ComfyUI Manager → "Comfyui_Comfly" → restart. No models download - it's an HTTP client. Key from https://ai.comfly.chat in Comflyapi.json (or the node's apikey).

Gotchas

Honest confusion risk: Pro and Max look like the same node with different defaults (Pro = png output, Max = jpeg), because the tier difference is on BFL's side, not in the interface. If you're not sure which to use, you're not alone - but the practical rule is: Pro when you need consistency and clean output for a pipeline, Max when you want the best single image and are willing to pay for it. And with no guidance/steps inputs, don't drag a Flex workflow over expecting your sliders to survive - they won't, and that's by design. One more thing worth repeating for this pack generally: check the response output when a result surprises you; the relay keeps task logs at ai.comfly.chat that say what the API actually did.