LayersFromBoundingBoxes - ComfyUI Built-in Node Documentation

This node converts an image batch and its bounding boxes into a layer stack, creating one layer per frame and placing each layer according to its matching box. Use it when a node outputs layers as a batch, because a batch carries only a single placement for every frame and the individual positions would otherwise be lost.

Inputs

Parameter	Description	Data Type	Required	Range
`image`	Image batch; each frame becomes one layer.	IMAGE	Yes	—
`bboxes`	Placement boxes, index-aligned with the image batch. Accepts bounding boxes (x, y, width, height), normalized elements (with a “bbox” — these need canvas_width/canvas_height to resolve to pixels), or a JSON string of either. Frames without a matching box are placed at the origin. A box’s width/height scales the layer to fit it. metadata.name (or desc) and metadata.z_index are used when present, and metadata.content_rect (frame-relative) crops the frame to its real content.	BOUNDING_BOX, ARRAY, or STRING	Yes	—
`mask`	Per-frame transparency, index-aligned with the image batch (1 = transparent, LoadImage convention).	MASK	No	—
`layers`	Layer stack to append to. Leave unconnected to start a new stack.	LAYERS	No	—
`crop_to_content`	Crop each frame to metadata.content_rect where present and place the content at the box position plus the rect offset. Leave on for batches whose frames are padded — it keeps only the real content at its true spot. (default: true)	BOOLEAN	No	true false
`canvas_width`	Document canvas width. 0 derives it from the placed layers. (default: 0)	INT	No	0 to MAX_RESOLUTION
`canvas_height`	Document canvas height. 0 derives it from the placed layers. (default: 0)	INT	No	0 to MAX_RESOLUTION

Notes:

bboxes and mask must be index-aligned with image: the nth box and the nth mask frame correspond to the nth image frame. Frames without a matching box are placed at the origin.
When bboxes contains normalized elements (with a “bbox”), canvas_width and canvas_height must be provided so those normalized positions can be resolved to pixels.
canvas_width and canvas_height must both be greater than 0 to set the document canvas explicitly. If either is 0, the canvas is derived from the placed layers or inherited from the connected layers stack.
When layers is connected, new layers are appended to it and receive z-index values above the highest z-index already present in the stack.
When crop_to_content is enabled and a frame has a metadata.content_rect, the frame is cropped to that rect and the box’s width/height scaling is not applied; instead, the rect’s offset is added to the box position.

Outputs

Output Name	Description	Data Type
`LAYERS`	The layer stack, ready for Create Layered Image.	LAYERS

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): a70956bf0d7ea8bdbd16767ed8b19600b274a6eeb745728f95219578adc73712

​Inputs

​Outputs

Inputs

Outputs