ComfyUI-ImageToLayers

ComfyUI-ImageToLayers

Separate any image (including AI-generated ones) into editable layers inside ComfyUI, using Image to Layers "Detailed Layers": the AI finds the editable elements, returns each as its own transparent layer, and rebuilds the background behind them.

Nodes

| Node | What it does | | --- | --- | | Image to Layers (Detailed) | Sends the image, waits for the job, returns the layers bottom to top. Optional parts_to_separate ("bottle, leaves, podium"). | | Image to Layers: Pick Layer | Takes layer index out of the batch (0 is the background). | | Image to Layers: Account | Shows who is signed in and the credits left; can sign out. |

Outputs of Detailed:

• layers – IMAGE batch, one image per layer, all the size of the background layer (the layers sit where they were in the picture).
• masks – MASK batch, the alpha of each layer (1 = the layer's pixels; ComfyUI's JoinImageWithAlpha expects the inverse for LoadImage masks, so use Pick Layer → JoinImageWithAlpha as in the example workflow to get a transparent PNG).
• layer_names – JSON list of the layer names, same order.
• job_id – the job on imagetolayers.com.

Install

ComfyUI Manager → search Image to Layers, or:

cd ComfyUI/custom_nodes
git clone https://github.com/jamesknightzhou-arch/ComfyUI-ImageToLayers
pip install -r ComfyUI-ImageToLayers/requirements.txt

Restart ComfyUI.

Sign in (no API key)

The first time you run the node it stops with a message containing a link and a short code. Open the link, check the code, approve (email or Google, in your browser), then queue the workflow again. The node picks the approval up by itself.

The session is stored on the machine that runs ComfyUI in ~/.imagetolayers/comfyui_session.json (mode 600) and refreshes itself. Use Image to Layers: Account with Sign out to remove it. No password or key is ever stored.

Credits and privacy

• Each run uses credits from your Image to Layers account, the same as Detailed Layers on the website.
• The image you connect is uploaded to imagetolayers.com for processing. Do not connect images you do not have the right to process.

Example

workflow_image_to_layers.json loads an image, separates it, and saves layer 1 as a transparent PNG (ComfyUI API format).

Settings (optional environment variables)

| Variable | Purpose | | --- | --- | | IMAGETOLAYERS_URL | Use another server, for example a staging site. Default https://www.imagetolayers.com. | | IMAGETOLAYERS_SESSION_FILE | Where the sign-in is stored. |

Develop

python3 -m unittest discover -s tests   # client and layout; no ComfyUI or torch needed

Limits

• Image up to 25 MB; the optional prompt is cut at 1000 characters.
• One image per run (the first image of a batch). A job can take a few minutes; the node waits up to 10 minutes.

License

MIT, see LICENSE. Using the nodes needs an Image to Layers account; the service itself is not covered by this license.