On a held-out test set of 50 identities (not seen during training):
In simple terms, V3 is a pipeline node (commonly used in ComfyUI) that extracts facial features from a source image and injects them into a target generation during the sampling process. face injector v3 work