Feature Request or (bug report?) - Allow embeddings inputs for FaceID node

[Mozoloa]

Hello !

I've been tinkering with the new nodes and am excited about the embedding combination nodes and especially the save embed node, that said, I mostly use FaceID and there's no embedding input on the node. So I always have to load all the pictures from my face or somebody else's face everytime I want to work on a workflow which is kinda annoying (I'm improving my workflow daily so I can't have one workflow per person, it'd be hell to maintain).

I get that insightface may need actual pixels to work in faceID case on top of the embeddings, but I wondered if there was a way I could actually save all of the input images in a nifty latent way as to load them easily like it's already possible with the new embeddings nodes.

I hope I'm clear, if not, don't hesitate to ask !

[cubiq]

At the moment is not possible to save the face embeds, but it's planned.

in the meantime you can save the images and batches as template or you can use a load image from path node that will load all images from a directory

[Mozoloa]

Hey thanks for your answer, I've gone the dir to batch way it works good enough !

Aren't face embeddings IPA embeddings + insightface embeddings ? I'm currently using reactor a lot and they have a blend face node and a save face (as safetensor) available that I use a lot. Maybe there's some idea to steal there !

Anyway I don't know much about it, just suggesting, have a nice day

[cubiq]

it's a bit more complicated because we have multiple embeddings for faceid and they are not compatible with each other, but it's doable just need to find the time

[GigaDroid]

Hey there, just wanted to check if there is any progress on this? I would need this as well @cubiq