zhenzhen_Flux_2_Max

Flux 2 Max is the top-quality tier of Black Forest Labs' image API, and Comfly_Flux_2_Max is the Comfly pack's paid client for it. The first thing you'll notice next to its sibling Comfly_Flux_2_Flex: there are no sampling knobs. No guidance, no steps, no prompt-upsampling toggle. That's not a bug - Max is the "just generate it as well as you can" tier, so the node is deliberately minimal. Your job is the prompt (and optionally the reference images); BFL handles the rest. As with everything in this pack, nothing runs on your machine - you're paying per image through the Comfly relay.

Inputs that matter

• prompt - required, multiline. Do your prompt-crafting here, because there's no upsampling helper on this node to fix a thin one.
• input_image … input_image_8 - up to eight reference images. Feed one in and the prompt becomes an edit/reimagination; stack several for character/style blending. The real power of the node is here.
• width / height - 64–6000, step 8, default 1024 square. Up to 6000 is genuinely useful for print-scale work on Max.
• safety_tolerance - 0–5, default 2. Higher = stricter content filtering. It's an API-model dial; the default is the sane place to be.
• output_format - jpeg (default) or png. Max defaults to jpeg because that's what you usually want for a final image; switch to png if you'll composite or re-edit.
• seed - -1 for random, or fix it to reproduce a result.

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

Install

Standard Comfly:

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

Or ComfyUI Manager → "Comfyui_Comfly" → restart. No models download; credits + key at https://ai.comfly.chat (Comflyapi.json or the node's apikey).

Gotchas

Max is the expensive tier of the three, and there's a real temptation to treat it as the default. Don't - iterate on Flex (or cheaper) until you've locked the composition, then spend Max's premium on the final. Also, the three Flux nodes being visually near-identical is a genuine trap: this one has no guidance/steps inputs, so if you swap it into a Flex workflow those fields just vanish from the graph. And remember the reference-image setup is the whole value of the multi-input design - a Max call with no references is just an expensive text-to-image.