Using multiple face_models

[squallseeker]

Are multiple face_model for a single node process possible? I've replaced multiple faces using a source image with multiple faces and indexing works well, but I've not figured out how to load (or concatenate) multiple saved face_models to feed the same reactor processing node; for those (when I'm using saved face_models), I have added second or multiple reactor nodes, one for each face model being used.

As a second comment, when there are multiple destinations there is often a point when the swap will overlay an undesired target. With a single image this may just mean freezing the last seed, updating an index or approach parameter and redoing, but if other changes are made or multiple items queued the only option is to open the pre-save and process it separately. Video can be quite problematic as some faces may appear briefly and receive temporary targeting. For either of these cases, a perhaps nice to have would be some form of target modeling. Since a specific checkpoint will often produce similar looking main focus characters, when there is more than 1 destination face detected, check to see which is the closest to matching a recent destination reference model (or tracking the position of prior swaps in the case of video image sequence and targeting the same destination closest to it).

If there were the ability to pause comfyui, just automate a pause condition to click the desired destination mask and continue might be possible, but that seems an unlikely add for comfy in the short term, so thinking of other options.

This is a terrific node set; thanks for it and your other tools!