What features do you most want us to add ?

[ResearcherXman]

We are a small team and cannot handle all feature requests at once. Thus, please tell us what is your favorite feature you wanna us add. Comment under this post. Meanwhile, we encourage all kinds of PRs (fixing bugs, adding new features, etc.) from the community, if you want to take on a feature, please open a new issue naming "Support for xxx" and we will assign you. Thanks for your interests, love you all ❤️.

[brunneis]

It would be great to support non-human faces. It would be great to add better support for non-human faces. Congrats for the great job!

Edit: it works on animals, but with many pictures I get a "Cannot find any input face image" error:

Traceback (most recent call last):
File "/root/.pyenv/versions/3.11.7/lib/python3.11/site-packages/cog/server/worker.py", line 217, in _predict
result = predict(**payload)
^^^^^^^^^^^^^^^^^^
File "/src/predict.py", line 173, in predict
face_info = sorted(
^^^^^^^
IndexError: list index out of range

[zewolf5]

Not sure if its already there, but multiple reference images? I could not find a way to select more than one image in the sample code provided. Topic #33 mentioned use of multiple reference images for better results.

Also is it possible to do some "adetailer magic" like upscaling the part of the image that contains the face, apply hires face-swap and then resize down again to the original image to get a detailed face when not covering the whole image?

I am still new to this, so ignore anything i have misunderstood.

[fofr]

I'd love to see implementations of each of the examples in the paper.

[tgohblio]

SD 1.5 support.

Opens up whole new possibilities with LORAs, AnimateDiff (as of today, it's SD 1.5 only since XL is in beta) and community-trained SD1.5 models.
Hint: this will drive more Github ⭐️

[nosiu]

Not sure if its already there, but multiple reference images? I could not find a way to select more than one image in the sample code provided. Topic #33 mentioned use of multiple reference images for better results.

I'm no expert but this should work
image = pipe( ... image_embeds=(face_emb1 + face_emb2 + face_emb3 + face_emb4)/4, ...)

[johndpope]

Downstream has some nice illustrations- https://github.com/ZHO-ZHO-ZHO/ComfyUI-InstantID -
I played around with thie workflows - but couldn’t get results that I was after. It seems all the examples are based off text concatenation. - somehow interrogating image for words to add to prompt
Said another way - what I want is - given an image - can instantid make new images that look like it? Or is this not a suitable use case?
I realise t2i adapter colour will best fit here - but then how does instantid shine? Can you create something that looks like this?

(this is the example in ComfyUI-InstantID)

[bent1e]

Adapt with IP Adapter to control the style of images by inputting ref instead of text prompt.

[johndpope]

Thanks @bent1e - looks like the ref image has been updated. testing now

[marsplant]

Face swapper is wanted!

[Kuvshin8]

Samplers? Sde/2M Karras or Euler A? I don't find any mentions about it

P.S. Thanks for your models

[brausesa]

Please make it possible for the local version to specify long tasks for generating images.... For example, so that I could specify one reference image of a face, plus specify several images of poses, plus, for example, select twenty different styles (prompts) and the number of variants (generations) for each style... and after pressing the "submit" button, I could go to sleep... And in the morning look ouput-folder with a hundred results :)

Thanks! You have made an incredible product!