Skip to content
BytePlusBytePlusBytePlus ModelArk for ComfyUI

Seedream Layer Separation

Template
Seedream Layer Separation, in the ComfyUI template browser under BytePlus
Workflow file
Seedream Layer Separation.json
BytePlus nodes
Seedream 5.0 Layer Separation
Keys needed
ModelArk API key
Also uses
Load Image, Preview Image, Save Image (built into ComfyUI)

What it makes

Section titled “What it makes”

A background image and a stack of layers. Seedream looks at one picture (the ComfyUI logo) and decomposes it into a base image plus the elements on top of it, each as its own layer. The example contains two layers: the yellow ring and the word ComfyUI. You can then move or edit each layer on its own.

The workflow

Section titled “The workflow”
Load Image feeding the Seedream Layer Separation node, whose base_image goes to Preview Image and layers go to Save Image
Four nodes: Load Image, Layer Separation, Preview Image and Save Image.
  1. Load Image supplies one picture. The node takes exactly one image.
  2. Seedream 5.0 Layer Separation runs with the model seedream 5.0 pro, an empty prompt (it finds the main elements by itself), size set to auto and crop_layers set to full canvas.
  3. Preview Image shows base_image, the background plate.
  4. Save Image receives layers and writes each layer to the output folder as BytePlus/Layers/Seedream. With full canvas every layer sits on a black canvas the size of the base at its own position.

Run it

Section titled “Run it”
  1. Add your ModelArk API key in Settings → BytePlus if you have not yet (API keys).
  2. Open the template browser and choose Seedream Layer Separation under BytePlus.
  3. Press Run. The background shows in Preview Image and the layers in Save Image.

Settings worth changing

Section titled “Settings worth changing”
  • Image: pick your own picture in Load Image. It must be one picture of at least 262,144 pixels (for example 512x512).
  • prompt: leave it empty to separate all major elements, or name the elements you want in plain words. You can also target regions with <bbox>left top right bottom</bbox> tags.
  • crop_layers: switch to minimal size to crop each layer to its bounding box. That gives smaller images; use the bboxes output to put them back.
  • save_layers: an advanced input. Turn it on to also write the base image and the original layer PNGs from the node itself.
  • prompt_optimization: standard is set; fast shortens generation time.

Example result

Section titled “Example result”
The purple gradient base image in Preview Image, and two saved layers on black: the word ComfyUI and a yellow ring
Example result: the background and two layers.

Time and cost

Section titled “Time and cost”

The node bills once per run through your ModelArk account. The seed control is set to randomize, so every run asks again and bills again. Set it to fixed and ComfyUI reuses the cached result of an unchanged workflow without billing.

Section titled “Related”