WeVisOne – AI-Powered Masking and Image Editing
This workflow uses the WeVisOne text encoder. ⚠️ Important: For now, only the 4B variant is supported.
How to use it
Load your image.
Tell WeVisOne what you want it to find or mask.
Adjust your editing prompt to describe the changes you want.
Hit Queue and let it do its thing!
The workflow can also be used with the LanPaint custom node, if you have it installed.
What makes WeVisOne interesting?
WeVisOne has a lot more to offer than just segmentation. It supports several visual tasks, including:
Segmentation: Create precise masks around specific objects or areas.
Bounding boxes: Detect objects and return their locations.
Point detection: Identify specific points within an image.
Background removal: Isolate subjects from their backgrounds.
WeVisOne could be a great addition to image-editing workflows, especially when you need accurate masks for specific regions.
It's a relatively small 4B model, so don't expect miracles, but its capabilities make it worth experimenting with!
Works with MOSS and KREA2
Downloads : https://huggingface.co/Comfy-Org/WeVisOne-VL/tree/main/text_encoders
Custom Node: https://huggingface.co/paralaif/ComfyUI-WeVisOne-VL