PixelPanda: AI Enhance

PixelPanda: AI Enhance is the pack's "make this look better" button, and the honest thing to know up front is that it's vague on purpose. No sliders, no model picker, no before/after preview. You feed it an image, it calls the PixelPanda API's /enhance endpoint, and a moment later you get an IMAGE back that has been deblurred, cleaned up, and generally given a bit more life. One credit per image.

What "enhance" actually means here

This is the "more detail" end of the upscaling spectrum, not the "more pixels" end. The KB's upscaling doc splits the field exactly there: a plain ESRGAN model or Lanczos makes the image bigger without inventing anything, while a generative enhancer adds plausible detail that was never in the source - sharpening soft edges, lifting texture, unblurring. The cloud backend behind this node is the second kind, which is why it's the right call for a soft, low-res, or lightly damaged image and overkill (and a paid overkill) for one that's already sharp and just needs to be bigger.

It's also worth saying what it is not: it is not a sharpen filter. A deterministic unsharp-mask pass runs in milliseconds for free and never hallucinates; the post-processing KB is emphatic that you don't burn a generative pass on work a lookup table does perfectly. Use Enhance when the image is genuinely soft or has no detail to recover - the case where a sharpen filter just amplifies noise.

The inputs

This node is about as close to a one-button operation as ComfyUI gets:

• image - what you're enhancing.
• api_key - a pk_live_… key, or blank to fall back on the PIXELPANDA_API_KEY environment variable.

That's the whole list. No strength, no scale, no mode. The output is a single IMAGE with the same dimensions as the input, ready for the next node in your chain.

Install and key

cd ComfyUI/custom_nodes
git clone https://github.com/RyanKramer/ComfyUI-PixelPanda

Restart ComfyUI and the node appears under image/PixelPanda. ComfyUI Manager users can just search "PixelPanda". There are no dependencies to install - the pack declares zero, using only the torch/numpy/Pillow that ship with ComfyUI. The key comes from https://pixelpanda.ai/profile and either goes in the node's field or in the PIXELPANDA_API_KEY env var before you launch. Enhance costs 1 credit per image; the $1.99 starter pack needs no subscription and credits don't expire.

Why you'd pay for this

The pitch is the same as the rest of the pack: a 1-credit API call instead of a GPU-hungry local model. Running a serious generative restorer like SeedVR2 or SUPIR locally is free after the electricity but wants real VRAM and some setup; this node wants none of either. If you're on a modest card, or you occasionally need to rescue a phone photo or an upscale source and don't want a second model in your life, that trade is defensible. If you're already running SeedVR2 in the same graph, this node is a strictly worse version of something you own - skip it.

Gotchas

Same family quirks as every node in this pack. It processes only the first frame of whatever batch you feed it, so don't expect frame-by-frame video enhancement. It needs a live internet connection and the service up; the call waits up to 300 seconds and prints [PixelPanda] /enhance ok - credits used/remaining to the console when done. Bad key gets you a clear 401 error, empty wallet a 402 - neither will leave you guessing. And since there's no "how much" control, if you need to tune the amount of restoration you'll want a local node instead; this one is all or nothing.