Nodes/PopNow/Shadow Remover · Edit
ComfyUI Node

Shadow Remover · Edit

Shadow Remover lifts the harsh shadow off a face or a product — check the label after

By popnow-ai·Created 24 days ago·Updated about 20 hours ago· 0
Shadow Remover · Edit
  • image_url
  • images
  • task id
◄aspect_ratiomodel default►
◄output_formatJPEG►
◄seed0►

This one is a retoucher's tool, not a generator's. If you shoot products on a table with one LED panel and no fill, you know the problem: a hard shadow under the bottle, a dark half on the model's face, a cast shadow that makes the whole frame look like it was lit by an intern. Fixing it properly is an hour in Photoshop with dodge and burn.

Shadow Remover · Edit is 4.5 credits and no settings. It is also, descriptively, the least controllable node in the pack - which is fine, because the job is narrow.

What it is

A cloud node over popnow/shadow-remover-v1/edit, in the PopNow pack. Photo in, de-shadowed photo out as a normal IMAGE. Runs on PopNow's servers, so no GPU and no fill-light checkpoint on your disk. The API-node pattern applies as always: cheap, fast, and your frame leaves the machine.

The inputs (there are barely any)

image_url - the photo, JPEG/PNG/WebP up to 10 MB. That's the only required thing and effectively the only decision you make.

aspect_ratio defaults to model default, so framing is preserved. Don't touch it; a de-shadowing pass that also re-crops is a bug report waiting to happen.

output_format defaults to JPEG. For product work going into a layout, PNG is worth the size.

seed starts fixed so the same graph twice gives you the same picture instead of a second charge. Randomize it if a result is nearly right and you want a variation, at 4.5 credits a roll.

Outputs: images and task id. No mask, no alpha - you don't get a shadow layer back, just a flatter picture.

What it's good at, and what it's not

Good: lifting a harsh cast shadow off a plain surface, opening up the shadowed side of a face, rescuing phone snapshots of things you're selling. The kind of correction where the "before" is obviously worse than the "after" and nobody needs to know it happened.

Not good: creative control. There's no prompt, no strength dial, no "keep the rim light" instruction. You can't say how much shadow to remove, and you can't ask for directional fill. That's the difference between this and a properly-described relight pass, where a sentence like "warm golden light from a window on the left" gives you a lever - the KB's IC-Light doc is a useful read on how relighting moved from dials to sentences, and why sentences beat dials for this family of jobs.

The local alternative is a masked fill or an instruction editor with a relight LoRA, which costs you an afternoon and gives you all the control. This node is for when you have forty SKU photos and no afternoon.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/popnow-ai/popnow-comfyui

Or Manager → search PopNow → Install → restart. Add a key from popnow.ai/api/console/keys in Settings > PopNow, or on a server:

POPNOW_API_KEY=sk-popnow-... python main.py

Nothing to download, no dependencies beyond what ComfyUI ships. Restart again after adding the key so the node list builds from your account's catalog.

Where people get burned

It repaints, so lettering will change. The model doesn't have a "shadows only" layer to edit - it re-renders the region it thinks is shadowed. Packaging copy, fine print, embroidery and small logos come back slightly different, which is a problem when the label is the product. Zoom to 100% on any text before you ship.

Removing a shadow can flatten the object. A product with no contact shadow looks pasted into the frame. Often the right move is a partial fix - or adding a soft shadow back afterwards, which is cheaper than asking this node to be subtle, since it has no strength setting.

It's not a colour corrector. If the real problem is white balance or exposure, this node isn't fixing it, and a de-shadowed but wrongly-white-balanced photo is still wrong. The post-processing notes are the better starting point for that.

Batches truncate. One image per image_url socket - connect a batch and only the first is uploaded (the pack logs that it dropped the rest). Queue a folder as a loop, and keep an eye on the credits: forty SKUs is 180 credits, about eighteen cents, which is still cheaper than your time.

CategoryPopNow/Image/Tools

Inputs (4)

NameTypeDefaultDescription
image_urlIMAGEThe photo to process: JPEG, PNG or WebP, up to 10 MB.
aspect_ratioCOMBOmodel defaultOptional output aspect ratio. Leave it empty to keep the shape of the photo.
output_formatCOMBOJPEGFile format of the returned image.
seedINT00–2147483647Same seed and settings, same result. Set 'control after generate' to randomize to get a new result on every run; every run is billed.

Outputs (2)

NameTypeDescription
imagesIMAGE—
task idSTRINGThe PopNow task id, for the API console logs or support.