Feature Request: Support for Multiple Faces with Unique Embedding IDs

[sg-byo]

Current Behavior:
The current sample only extracts the maximum face from the input image and uses it as the embedding.

Requested Feature:
Is there a plan to support multiple faces, each with a unique embedding ID or index?

Expected Behavior:

Option 1: The model should support the use of a unique embedding ID for each face, which can then be included as part of the prompt. Similar to how Lora embeddings are used.
Option 2: Allow the user to provide multiple indexed face images, with each face having its own unique embedding.

This feature would enhance the flexibility of the system, enabling better handling of images containing multiple faces.

[EndlessSora]

Thanks for your suggestion. We may look into this when our bandwidth is sufficient.

[GraftingRayman]

this can be done by editing the facexlib files, check my pulid repository, i have added these features to my version of pulid

[Fathaah]

Thanks for your suggestion. We may look into this when our bandwidth is sufficient.

I tried tinkering a bit with ComfyUI, made a node and was able to get multiple IDs somewhat working with Conditioning Masks. You can find the workflow here-

https://github.com/katalist-ai/ComfyUI-InfiniteYou

Let me know what you think about this. It works with aes projection weights and sim controlNet.

[EndlessSora]

Thanks for your suggestion. We may look into this when our bandwidth is sufficient.

I tried tinkering a bit with ComfyUI, made a node and was able to get multiple IDs somewhat working with Conditioning Masks. You can find the workflow here-

https://github.com/katalist-ai/ComfyUI-InfiniteYou

Let me know what you think about this. It works with aes projection weights and sim controlNet.

Thanks a lot for your contribution! We currently do not have a chance to test it but will consider linking great unofficial contributions. We are also planning an official ComfyUI.