photoshop_preset_Unpack

Unpacks whatever photoshop_preset_pack bundled - a canvas image, a mask, a slider value, a seed, positive/negative prompt text, and target width/height - back into eight individually-typed outputs you can wire straight into a normal generation pipeline.

Part of ComfyUI-Apt_Preset's stack category. No processing happens here, just a bundle coming back apart.

How it works

Feed it the ps_pack output from photoshop_preset_pack and it hands back all eight values exactly as they were packed, correctly typed this time - Canvas as an IMAGE, Mask as a MASK, Slider and Seed as FLOAT/INT, pos/neg as STRING, W/H as INT. From here they go wherever a normal workflow would send them: Canvas into an image loader or VAE Encode step, pos/neg into your CLIP Text Encode nodes, W/H into an Empty Latent Image, and so on. The bundle-then-unpack pattern exists so that setup lives in one compact node rather than eight scattered widgets, especially useful if you're saving this whole combination as a reusable preset.

It's also worth using even if you only care about two or three of the eight fields - you don't have to wire up every output. Leave the ones you're not using dangling; ComfyUI won't complain about an unused output socket the way it would about a missing required input.

Inputs and outputs

• ps_pack - the required FPIPE2 input, produced upstream by photoshop_preset_pack. Custom type, specific to that node.

Eight outputs: Canvas (IMAGE), Mask (MASK), Slider (FLOAT), Seed (INT), pos (STRING), neg (STRING), W (INT), H (INT).

Installing it

ComfyUI Manager: search "ComfyUI-Apt_Preset", install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/cardenluo/ComfyUI-Apt_Preset.git

Restart afterward. install.bat handles Windows dependencies; on Linux/Mac, install from the requirements file inside your ComfyUI venv. No model or external file dependency for this node.

Common issues

Nothing to plug into ps_pack. You need a photoshop_preset_pack node upstream - this input type only accepts that specific pack node's output, not a raw image or string.

Canvas or Mask comes out empty. Those two fields are optional on the pack side - if the upstream photoshop_preset_pack node never had an image or mask wired into Canvas/Mask, the unpack step has nothing to hand back and will pass through empty/default values rather than erroring.

pos/neg text came back blank even though you typed something upstream. Double-check you typed it into the photoshop_preset_pack node itself and not into a separate CLIP Text Encode you assumed was connected - these are independent text fields on the pack node, not automatically synced with anything else in your graph.

W/H don't match the actual size of Canvas. They're independent fields, not derived from the image - photoshop_preset_pack doesn't read the canvas dimensions and fill these in for you, so if you changed the image without also updating the width/height fields upstream, the two will disagree. Feed Canvas through a "Get Image Size" style node instead if you need dimensions that always match the actual image.