search

invoke-ai / InvokeAI

InvokeAI is a leading creative engine for Stable Diffusion models, empowering professionals, artists, and enthusiasts to generate and create visual media using the latest AI-driven technologies. The solution offers an industry leading WebUI, supports terminal use through a CLI, and serves as the foundation for multiple commercial products.
https://invoke-ai.github.io/InvokeAI/
Apache License 2.0
22.35k stars 2.32k forks source link

[enhancement]: [Controlnet] - IpAdapter Face - Multiple images input #6084

Open suonnon opened 3 months ago

suonnon commented 3 months ago

Is there an existing issue for this?

Contact Details

No response

What should this feature add?

Multiple images input would help the IpAdapter Face controlnet detect and generate the face more precisely. This is available in A1111. I hope Invoke would compile this feature into controlnet section soon!

Alternatives

No response

Additional Content

No response

psychedelicious commented 3 months ago

@RyanJDick @hipsterusername Isn't this the same as adding multiple IP Adapters with the face model in the linear UI?

suonnon commented 3 months ago

@RyanJDick @hipsterusername Isn't this the same as adding multiple IP Adapters with the face model in the linear UI?

Does multi IP-Adapters (one input image for earch Ip-Adapter) work the same as one IP-Adapter (with multi input image).?I'm not a dev person, so I don't know. But in A1111, multi images input with one Ip-Adapter works really well.

StellarBeing25 commented 3 months ago

Just wanted to add that the IP Adapter node in ComfyUI got updated with new features like setting embeds combine methods that would be only possible in invokeAI with multi-input support in IP Adapter. IPv2-1

RyanJDick commented 3 months ago

InvokeAI already supports:

  1. Multiple IP-Adapters (in both Linear UI and Workflows).
  2. Multiple images to an IP-Adapter (Workflows only).

These two features are not equivalent.

@suonnon It sounds like you may already be able to achieve what you want by using "multiple images to an IP-Adapter" in the Workflows UI. Let me know if that is sufficient, or if there is still something missing.

@StellarBeing25 We do not currently support different combine_embeds options - we always use concat when multiple images are passed to an IP-Adapter. Is there a particular workflow that you are trying to implement with this feature? If yes, it's probably worth creating a new Github issue to track that separately.