心宝❤合成PSD
A stack of generated images into a real layered Photoshop file
- images
- psd_path
- preview
Most ComfyUI saves are flat files: one image, one PNG. This node is the exception that makes sense when you're doing product work - it takes a batch of images and stacks them as layers in a real Photoshop .psd file. If your final product page is going to be edited in Photoshop anyway, having each generated screen arrive as its own layer instead of a folder full of PNGs is a genuine workflow win.
It's one of the pack's e-commerce utility nodes, and it leans on the psd-tools Python library (listed in the pack's requirements as psd-tools>=1.9.0) to write the file. Nothing about it touches a model or the network - pure local file assembly.
The inputs
images(IMAGE) - the batch of images to layer into the PSD. One batch = one file with N layers. Wire in the output of whatever generation node produced your screens; the layer order follows the batch order.filename_prefix(STRING, defaultpsd_batch) - the output name. Files land in ComfyUI's output directory, sopsd_batchbecomesoutput/psd_batch_00001.psd(counter appended). Name it per product and it doubles as a search key in your output folder.
The two outputs
psd_path(STRING) - the absolute or relative path of the written file. Handy if you're building an automation that needs to know where the PSD went without globbing the output folder.preview(IMAGE) - a flattened preview of the PSD, so you can eyeball the composite in the ComfyUI UI without opening Photoshop. Wire it to a preview node and you get a visual "this is what the layered file looks like merged."
When you'd reach for it
The pack pairs this with its detail-page workflow: generate N screens for a product listing, then stack them into one PSD where each screen is a layer. There's also a companion utility, 心宝❤分层数 (XinbaoLayerSelect), whose only job is to pick how many layers you want - you can wire its INT output into the batch count upstream so the whole "how many screens / how many layers" question is one dropdown.
If your downstream is print or an editor that wants layers, this is the node. If you're going straight from ComfyUI to a website, it's overhead - a flat grid export serves you better. The PSD is only worth its weight when someone's actually going to open it.
Installing
Ships in comfyui-banana-li-linux:
cd ComfyUI/custom_nodes
git clone https://github.com/98624017/comfyui-banana-li-linux
Restart ComfyUI. This is the one node in the pack where psd-tools is a hard requirement - if the node fails to import, that's the missing piece (pip install "psd-tools>=1.9.0" in your ComfyUI Python environment). The README's dependency list covers it, but it's easy to miss if you installed the pack's other deps selectively.
Two gotchas. First, layer count = batch size, so a 1-image "batch" gives you a one-layer PSD - check your batch upstream if the file seems thin. Second, PSDs with many full-resolution layers get large fast; at 2K per layer, a 20-screen listing is a file that'll make Git and email both unhappy. Flatten copies for delivery, keep the layered master local.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | — | |
| filename_prefix | STRING | psd_batch | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| psd_path | STRING | — |
| preview | IMAGE | — |