ComfyUI Flux Redux + PulID Workflow | Perfect Face Consistency in Any Style | No More Errors

Поділитися
Вставка
  • Опубліковано 26 гру 2024

КОМЕНТАРІ • 30

  • @noNumber2Sherlock
    @noNumber2Sherlock 28 днів тому

    Great video and have the workflow all working but the end result looks just like the source image. Meaning if I followed your example, the output does not look like Anne but stays the same as the image I wanted to put Anne's face on but just looks the same as the target image albeit changes a little bit.

    • @RikkTheGaijin
      @RikkTheGaijin 27 днів тому +1

      this is becasue this video is fake and it's wasting everyone's time. You cannot trasfer a likeness with Redux. It's just not possible.

    • @edwardtse8631
      @edwardtse8631 25 днів тому +1

      @@RikkTheGaijin It is not fake, I got it to work, just ask why it is not working for you .

  • @nickfai9301
    @nickfai9301 27 днів тому +1

    Thank you. I like your toturial because you explain each nodes one by one. Really suitable for a beginner like me.

  • @reitani-l4q
    @reitani-l4q 15 днів тому

    Thank you for sharing this workflow. I tried it and it worked very well.
    But if I simply want to change the face of a photo of, let’s say, a female celebrity wearing a beautiful dress and I want to swap her face with mine.
    I tried using flux fill inpainting and pulid workflow and it could run without error, but the come-out picture is nothing to do with my face.
    Do you know something about this?
    I think if it worked, it should be the simplest workflow to swap faces.
    Maybe you have a better method to do this? Please share it with me. Thank you a lot.

  • @rauanmukhametkan1940
    @rauanmukhametkan1940 13 днів тому

    hello, can you help with this error ? forward_orig() takes from 7 to 9 positional arguments but 10 were given

  • @zengrath
    @zengrath 21 день тому

    Where do you suggest i add a lora node in this workflow if i need a lora to help with the generation.
    throwing a normal lora loader between the diffusor and clip loader and running outputs to encoder/pulid/text nodes seem to be doing okay for me. lora is helping to get what i was looking for.

    • @ComfyUIworkflows
      @ComfyUIworkflows  20 днів тому +1

      I will update the workflow with LoRA. Once it's done, I will update you regarding workflow

  • @yapyh2872
    @yapyh2872 26 днів тому

    No matter how i try, i'm getting the "expected scalar type Half but found BFloat16" error"

    • @ComfyUIworkflows
      @ComfyUIworkflows  26 днів тому

      try to update your comfyui & you have a old graphic which pulid not support

  • @andreabucci5819
    @andreabucci5819 16 днів тому

    insightface-0.7.3-cp39-cp39-win_amd64.whl is not a supported wheel on this platform.

  • @miriyalapraveenkumar9160
    @miriyalapraveenkumar9160 28 днів тому +1

    hi pip install -r requirements is getting below error i tried to delete and add requirements.txt file using pip freeze > requirements.txt commend but still getting same error
    Could not open requirements file: [Errno 2] No such file or directory: 'requirements'

    • @ComfyUIworkflows
      @ComfyUIworkflows  28 днів тому

      In ComfyUI Inside custom node folder. look for the pulid_flux folder. open that folder in cmd then type pip install -r requirements.txt

    • @miriyalapraveenkumar9160
      @miriyalapraveenkumar9160 28 днів тому

      @@ComfyUIworkflows yeah i tried it. after that getting like this means while starting comfyui.
      (IMPORT FAILED): C:\Drive_1\AI\ComfyUI\App\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux

    • @miriyalapraveenkumar9160
      @miriyalapraveenkumar9160 28 днів тому

      facexlib missing

    • @miriyalapraveenkumar9160
      @miriyalapraveenkumar9160 28 днів тому +1

      installed successfully i used this in my python environment "python -m pip install --use-pep517 facexlib"

    • @TheSslawek89
      @TheSslawek89 28 днів тому

      @@miriyalapraveenkumar9160 yea thats working

  • @fixelheimer3726
    @fixelheimer3726 28 днів тому

    What if the face in the end result looks too photographic compared to the style of the image (comic)/base image for style ?

  • @占酷
    @占酷 27 днів тому

    `fa2F@v2.6.3-24-gbdf733b` is not supported because:
    requires device with capability > (8, 0) but your GPU has capability (7, 5) (too old)
    bf16 is only supported on A100+ GPUs
    `cutlassF-pt` is not supported because:
    bf16 is only supported on A100+ GPUs

    • @yapyh2872
      @yapyh2872 26 днів тому

      So means old GPU can't run Pulid? no wonder i get the error "expected scalar type Half but found BFloat16" whenever it runs till about 80%

    • @ComfyUIworkflows
      @ComfyUIworkflows  26 днів тому

      you need to upgrade you graphic card

  • @RikkTheGaijin
    @RikkTheGaijin 27 днів тому +2

    This video is fake and it's wasting everyone's time. You cannot trasfer a likeness with Redux. It's just not possible.

    • @Mattyooooooooooooooo
      @Mattyooooooooooooooo 27 днів тому

      Not possible with Redux specifically, but as a 2pass with the second pass through flux with Pulid or faceswap would be plausible. But only in specific conditions. This is how people are bypassing the API pay wall of the prompt use with redux by averaging prompt conditioning through another model. In any case, it's possible but maybe not the best practice.

    • @ComfyUIworkflows
      @ComfyUIworkflows  27 днів тому +1

      with the help of PuLid it possible, you can try yourself

    • @Dunc4n1d4h0
      @Dunc4n1d4h0 27 днів тому

      Sorry, but you don't understand how it works.

  • @edwardtse8631
    @edwardtse8631 26 днів тому

    After 2 hours of struggle, I finally got it to work. For some reason antelopev2 models are extracted in the wrong directory. ComfyUI\models\insightface\models\antelopev2\xxxx are extracted to ComfyUI\models\insightface\models\antelopev2\antelopev2\xxxxxx , that extra antelopev2 directory gave me errors. After that, everything works for me. I am able to have a source file, with a clown face and merge the clown face. For people who said this completely not work, this is so not true, it works if you get the workflow to work.