Modelverse Flux Kontext Pro

If Kontext Max is the flagship, Kontext Pro is the specialist - Black Forest Labs' model tuned for single-image editing and enhancement, served through UCloud's Modelverse API just like everything else in this pack. The interface is nearly identical to the Max node (same prompt, images, seed, guidance_scale, num_requests), but the model underneath is the one you want for "fix this one image really well" rather than "merge this whole batch." Pro is also the cheaper tier, which makes it the sensible default for everyday editing.

What it is

A single-image editor with the same multi-image trick as Max: wire one IMAGE and it edits; wire several (via Modelverse Image Packer) and it edits them as a batch. The README frames it as "expert single image editing and enhancement," and that's the honest job description. Style transfer, lighting fixes, adding/removing elements, upscaling the idea of an image - one prompt, one image in, one image out.

The inputs that matter

• client - required, from the Modelverse Client / Secret Client node.
• prompt - the edit instruction. Imperative, specific: "turn the sky into dusk," "make it a watercolor."
• images - required. A single IMAGE, or a batched tensor from the Image Packer for multi-image mode.
• guidance_scale - 2.5 default (0–10). Editing models reward restraint.
• seed - -1 for random, or lock it for reproducibility.
• num_requests - 1–10, the bill multiplier.

Output is a single image (IMAGE tensor), ready to preview or save.

Installing and keying up

The whole pack is one install. ComfyUI Manager → search "ComfyUl-UCloud" → Install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/ucloud/comfyui-plugin

restart ComfyUI. Light deps (requests, pillow, numpy, openai, GitPython), no model files - Kontext Pro runs on UCloud's hardware. Get a ModelVerse API key from the UCloud console and enter it via the Modelverse Secret Client node (secret name in the workflow, key in secrets.json) or a config.ini MODELVERSE_API_KEY=... line.

Common issues

• Multi-image mode doesn't engage - like Max, it's detected from the input tensor's shape. Feed it a single image and you're in single mode. For a batch, use the Image Packer.
• "Input image(s) is required" - prompt-only isn't allowed here. No image wired, no run.
• "API key is not set" - empty client node. The eternal first check.
• Edits feel less "creative" than Max - correct. Pro is the specialist for faithful edits; if you want wilder reimagining, that's Max's lane.
• Costs - Pro is cheaper than Max per run, but num_requests still multiplies. Draft, then spend.

Here's the division of labor worth internalizing: Pro for edits that must respect the source, Max for edits that can run free. Both speak the same interface, so switching between them is a drag-and-drop, not a rebuild.