RedNode Studio Advanced

A "finished" Krea 2 render isn't usually finished. You run a second sampler pass to clean it up, then a face detailer, maybe a third pass on the eyes - and in a normal ComfyUI graph that's a chain of KSamplers and detailer nodes wired together, each with its own steps, CFG and sampler, and a spiderweb of connections that's easy to misread and a pain to reorder. RedNode Studio Advanced replaces that spiderweb with a visual list: each post-render pass is a card in a stack, with its own rig, steps, CFG, sampler, scheduler and start/end step window, and there are no wires between the passes at all.

This is the "post-render passes" half of the RedNode Studio pack, and it's what makes the example Ultima workflow so small. The main render happens upstream - usually the workspace's image output - and this node just walks a list: sampler refine, SAM3 face detailer, another refine, in whatever order you've stacked them. Take the finished picture out the other end and put the grading chain after it.

How it works

The config input is a JSON pass list, and the node's panel edits it for you - you'll barely ever type JSON by hand. Each pass names a rig from the Models tab (which pulls in the model, clip, VAE and LoRA stack), a type (sampler refine or SAM3 detailer), and its own sampling numbers. An empty prompt box on a pass falls back to the rig's own Prompts-tab row, so you don't have to duplicate text everywhere.

Two mechanism details are worth knowing because they explain what you'll see. First, each pass inherits from the rig when it doesn't set its own number - a detailer pass picks up the rig's detailer_steps, a sampler pass its main steps - and an unknown sampler or scheduler degrades to euler/simple rather than failing the queue. Second, the SAM3 detailer passes need ComfyUI-Easy-Sam3 for their face, hair and hands masks. Without that pack, the pass doesn't crash: it prints a note and passes the picture through untouched, so the graph keeps running minus the detailer.

The outputs

Two outputs: image, the finished frame, and report, a string that narrates what each pass actually did - useful when a pass was skipped because a rig was missing a model, or because Easy-Sam3 wasn't installed. When the report says "pass skipped," that's the node being honest about a missing dependency, not a silent failure.

Install

Same pack, same command:

git clone https://github.com/RedNodeAI/ComfyUI-RedNodeStudio.git ComfyUI/custom_nodes/ComfyUI-RedNodeStudio

Restart, or search RedNode Studio in ComfyUI Manager. No pip dependencies. But if you want the face detailers to do anything, you also need ComfyUI-Easy-Sam3, and its sam3.pt checkpoint in ComfyUI/models/sams. ComfyUI Manager will offer the pack's extra dependencies when you load the Ultima example workflow; the sam3.pt file is yours to fetch either way. Without it the node still runs - it just hands the image straight through the detailer passes.

What to watch for

The passes run in order and each one is a full sampler load - a four-pass stack on a mid-range card can add real time per image, which is exactly why the pack ships RedNode Free VRAM to unload the main model before the next one loads. And because the node reads rigs from the workspace by discovery rather than by wire, remember the rig actually has to exist: a missing rig logs a skip and hands the image on, which is graceful but means your "detailer pass" quietly didn't happen. Read the report string, not just the image, and you'll catch that.