Sa2VA Video Segmentation

XJSa2VAVideoSegmentation

Sa2VA (Segment Anything 2 with Vision Assistant) for video/batch processing. This node generates consistent segmentation masks across multiple frames or images using text prompts, with optional morphological refinement for cleaner results. Ideal for video processing or batch image segmentation.

Pack: ComfyUI-Sa2VA-XJ

custom_nodes.ComfyUI-Sa2VA-XJ

Inputs (11)

Name	Type	Required
model_name	COMBO	required
images	IMAGE	required
segmentation_prompt	STRING	required
threshold	FLOAT	required
use_8bit	BOOLEAN	required
use_flash_attn	BOOLEAN	required
unload	BOOLEAN	required
morph	COMBO	required
erode_kernel	INT	required
dilate_kernel	INT	required
iterations	INT	required

Outputs (2)

Name	Type
text_output	STRING
masks	MASK