AUN KSampler PlusV2 (Deprecated)

Let's be straight about this one: AUNKSamplerPlusV2 is the older progressive two-pass sampler from the AUN pack, and the pack itself has moved on. It's superseded by KSampler Plus (AUNKSamplerPlusv3) and KSampler 2-Model (AUNKSamplerPlusv4), which are the same idea with the rough edges worked off. You'd only reach for V2 today if you're maintaining an old workflow that already uses it - and even then, the migration is worth it. The article you actually want is the v3 one; the mechanism here is the ancestor of it.

What it was

The pitch was the hi-res fix, packaged as one node. Instead of generating small, upscaling, and running a second KSampler with a low denoise - the classic two-pass dance that's still the correct answer for generating above native resolution - V2 tried to automate the whole chain: a base pass, a latent-space upscale, a second pass at the higher resolution, then an optional pixel-space upscale and an optional final refinement. That's why its outputs are Base image, Latent upscaled, Image upscaled, Both upscaled, Refined image, plus LATENT and an Upscaled type string you could drop into a filename to say which output you actually used.

The controls followed the same logic: steps_total / steps_first to split steps between passes, start_step_second to control the second pass start, latent_upscale + ratio + upscaling_denoise for the latent re-sample, and image_upscale / image_upscale_method / image_upscale_model / image_upscale_ratio for the pixel-space leg, with image_upscale_refine + img_refine_steps + img_refine_denoise for the optional final refinement. All of that survives, refined, in v3.

Why it's deprecated

Two reasons, essentially. It was the model-only sampler before the pack learned to do better, and the successor split the difference more cleanly - v3 cleans up the pass logic, and v4 adds the option to run the latent-upscale pass on a second model entirely. The pack labels it deprecated and says it'll be removed in a future release, which is the practical signal: keep it if a saved workflow needs it, don't start anything new on it.

Installing it

It ships in the AUN pack regardless, so if you load an old workflow you already have it:

• ComfyUI Manager: search "AUN ComfyUI Nodes", install, restart.
• Manual: cd custom_nodes && git clone https://github.com/loz2754/AUN-ComfyUI-Nodes, then restart. Manual installs hitting ModuleNotFoundError: cv2 need pip install -r custom_nodes/AUN-ComfyUI-Nodes/requirements.txt.

Common issues

The honest advice: don't troubleshoot V2 in depth - replace it with AUNKSamplerPlusv3, which accepts the same inputs in the same shape and is what the maintainer is actually supporting. When you swap, check the pass settings (start_step_second, upscaling_denoise) carry over, because the defaults differ slightly. And remember the second-pass denoise rule from the upscaling world: 0.3–0.5 keeps composition, 0.6+ starts changing the image.