Batch CLIPSeg (BatchCLIPSeg) - ComfyUI-KJNodes

The Batch CLIPSeg node technically utilizes a pre-trained CLIPSeg model to align input text prompts with image features and calculates the relevance score between each pixel and the text.

The Batch CLIPSeg node is an image segmentation tool based on the CLIP model, capable of automatically generating corresponding masks for an image or a batch of images based on text descriptions. Its core function is to transform natural language instructions into precise selections of visual regions, enabling semantically-driven mask creation.

Node Function

This node technically utilizes a pre-trained CLIPSeg model to align input text prompts with image features and calculates the relevance score between each pixel and the text. It supports processing batch inputs of the IMAGE type, consisting of single or multiple images. It converts the relevance score map into binary or continuous masks through thresholding and optionally applies post-processing such as Gaussian blur to the masks.

Node Parameter Description - Batch CLIPSeg

Connection Inputs (Inputs)

Parameter Name	Data Type	Required	Default Value	Value Range/Options	Description
`images`	IMAGE	Yes	-	-	images (Input Parameter): Input image(s) or image batch to be segmented. Supports single or multiple images.
`opt_model`	CLIPSEGMODEL	No	-	-	opt_model (Input Parameter): Optional external CLIPSeg model input, used to replace the default model loaded internally by the node.

Control Parameters (Parameters)

Parameter Name	Data Type	Required	Default Value	Value Range/Options	Description
`text`	STRING	Yes	-	-	"sky".
`threshold`	FLOAT	Yes	0.5	0.0 - 10.0 (Step: 0.001)	threshold (Input Parameter): Sets the confidence threshold for segmentation; regions above this value are retained. Range 0.0 - 10.0, step 0.001.
`binary_mask`	BOOLEAN	Yes	True	-	binary_mask (Input Parameter): Controls whether to binarize the mask (black and white). If false, outputs a continuous grayscale mask.
`combine_mask`	BOOLEAN	Yes	False	-	combine_mask (Input Parameter): Controls whether to combine all currently generated masks into a single mask.
`use_cuda`	BOOLEAN	Yes	True	-	use_cuda (Input Parameter): Controls whether to use CUDA (GPU) for accelerated computation to improve processing speed.
`blur_sigma`	FLOAT	No	0.0	0.0 - 100.0 (Step: 0.1)	-
`prev_mask`	MASK	No	None	-	-
`image_bg_level`	FLOAT	No	0.5	0.0 - 1.0 (Step: 0.01)	-
`invert`	BOOLEAN	No	False	-	-

Output

Parameter Name	Data Type	Description
Mask	MASK	Mask (Output Parameter): The output mask, which can be used by subsequent image processing nodes.
Image	IMAGE	Image (Output Parameter): The output image, typically the same as the input image, used for workflow continuity.

Usage Scenarios

In practical applications, this node is commonly used in workflows requiring automated matting based on object or scene descriptions. For example, when batch processing product images, you can input the text "a shoe" to quickly generate precise masks of the shoes for all images containing shoes, facilitating subsequent background replacement or local adjustments.

Notes

The node's processing speed is affected by the model size and whether GPU acceleration is used. For high-resolution images or large batches, processing time may be longer. Furthermore, segmentation accuracy largely depends on the accuracy of the text description and the coverage of the model's training data.

Batch CLIPSeg Node Source Code Link

The Batch CLIPSeg node is from the ComfyUI-KJNodes node package.

Batch CLIPSeg(BatchCLIPSeg)

Batch CLIP Seg