Best Face ID workflow - how to use for multi characters/faces?
if multi face is not supported, is it possible to supply 1 face to BestFaceID lora, but it will only apply that face reference to only 1 of the 2 characters in the image input?
I'll look at multi-reference increase on it and a multiple person detector. I think it can be worked into it using the multi-reference IC Lora on top.
Isn't the LoRA only trained on the ID 2? I think its possible to pass in other Id's like 3,4,5... and probably chain that LTX Identity Transfer but from what i read its only trained on Id 2 if im not mistaken
It all depends on if using more source_id will actually work like it should, the model itself isn't really set up to be used properly with multiple i2v references like other models do. I would expect it to start becoming confused and merging people and making transitions tbh.
Currently looks like if you do chain and add other IDs it basically looks like it's doing an overwrite, I don't think it would be as simple as detecting 2 bbox from a single image, they would probably bleed together because the entire image is encoded as a single latent,