ComfyUI Nodes
Install ComfyUI-PixelPanda and five nodes appear under image/PixelPanda. They call the PixelPanda API, so there is no extra checkpoint to download and no VRAM taken from your generation graph.
Post-Processing That Does Not Compete for VRAM
Most ComfyUI clean-up nodes load another model into the same GPU your generation graph is already filling. These do not — the work happens in the cloud and the tensor comes straight back.
Generate as usual, then send the latent output through PixelPanda for the finishing steps: a clean cutout, a 4x or 8x upscale that holds up in print, an enhancement pass, text removal, or a prompt-driven edit with FLUX Kontext Pro. Each node takes an IMAGE tensor and returns an IMAGE tensor, so it drops into an existing graph without adapters or format juggling.
The five nodes
- Remove Background — clean cutout, alpha composited for downstream nodes (1 credit).
- AI Upscale — 2x and 4x for 1 credit, 8x multi-pass for 2, in fast, balanced or high-detail modes.
- AI Enhance — sharpen, deblur and fix colour on a finished render (1 credit).
- Remove Text/Watermark — wipe stamps and overlays from source images (1 credit).
- Edit With Prompt — FLUX Kontext Pro text-guided edits with a strength slider (2 credits).
Where It Fits a ComfyUI Graph
The finishing steps, without another checkpoint.
No VRAM cost
Processing runs in the cloud, so your GPU stays dedicated to generation.
Upscale to 8x
Multi-pass AI super-resolution for print-ready output from a standard render.
Clean cutouts
Remove Background returns an alpha-composited image ready for the next node.
Prompt editing
FLUX Kontext Pro edits a finished image from a text description with a strength slider.
No extra dependencies
Pure Python on top of the torch, numpy and Pillow ComfyUI already ships.
IMAGE in, IMAGE out
Standard tensors both ways, so the nodes slot into any existing graph.
Pure Python, no extra requirements
Uses the torch, numpy and Pillow already in ComfyUI. MIT licensed.
How It Works
Clone, key, drop the node in.
Install the nodes
Search PixelPanda in ComfyUI-Manager, or git clone ComfyUI-PixelPanda into ComfyUI/custom_nodes.
Add your API key
Set PIXELPANDA_API_KEY before launching ComfyUI, or paste the key into the node field.
Drop in a node
Add any node from the image/PixelPanda category and wire an IMAGE output into it.
Queue the prompt
The processed tensor returns to the graph and flows into Save Image or the next node.
Who Uses PixelPanda + ComfyUI
People whose renders have to become products.
Print-on-demand sellers
Upscale generated art 4x-8x in high-detail mode so it meets DPI requirements on shirts, posters and mugs.
AI artists
Cut out a subject cleanly for compositing without loading a segmentation model beside your checkpoint.
Product photographers
Run generated product renders through enhancement and background removal in the same graph.
Low-VRAM setups
Keep the whole GPU for generation while post-processing happens in the cloud.
Batch pipelines
Queue hundreds of images and let each one finish through the same node chain unattended.
Prompt-based retouching
Edit With Prompt applies a described change to a finished render without rebuilding the graph.
Frequently Asked Questions
How do I install the PixelPanda ComfyUI nodes?
Do the nodes use my GPU or VRAM?
Which nodes are included?
How much does it cost?
Where do I put my API key?
Does background removal keep transparency?
Is this good for print-on-demand?
Add PixelPanda to ComfyUI
Search PixelPanda in ComfyUI-Manager, or clone it into custom_nodes.
Explore More Resources
Browse our guides, tutorials, and tools