ComfyUI: Consistent Character Using Flux Model

Поділитися
Вставка
  • Опубліковано 22 лис 2024

КОМЕНТАРІ • 61

  • @VairalKE
    @VairalKE 26 днів тому

    Amazing tutorial and very well articulated. Thank you.

  • @ragemax8852
    @ragemax8852 Місяць тому +1

    Thanks, my friend, as this was an awesome workflow. No doubt, I will be using it to generate new images from scratch for now on, as I love how it's enhanced with the face swap feature. I no longer have to use stuff like Rope or FaceFusion to swap faces out. I did have one question, though and that was if there was a way to incorporate image to image with this as well? I wanna be able to load and use images I created elsewhere and take advantage of the face swap feature.

    • @techzuhaib99
      @techzuhaib99  Місяць тому

      Thankyou. Offcourse you can use other images for faceswap. Just load them with load image node and feed into re_actor

  • @fernandoemmanuel9281
    @fernandoemmanuel9281 Місяць тому

    Perfect. Very well explained. It worked perfectly!
    Thank you very much!

  • @williamhetherington2422
    @williamhetherington2422 2 місяці тому

    great value, subbed with notifications on. look forward to learning something on the next 1

  • @youtubeccia9276
    @youtubeccia9276 Місяць тому

    excellent

  • @matvfx6860
    @matvfx6860 2 місяці тому

    Great tutorial, thank you for sharing!

  • @vedangdate
    @vedangdate Місяць тому

    amazing!

  • @seloslav
    @seloslav 10 годин тому

    I keep getting this error at the last step: "Warning torch.load doesn't support weights_only on this pytorch version, loading unsafely. !!! Exception during processing !!! unexpected EOF, expected 4827818 more bytes. The file might be corrupted."
    Looking online, people are saying to re-download GFPGANv1.3.pth facerestore_models. That doesn't seem to work. What do you suggest?

  • @JOmoikane
    @JOmoikane 2 місяці тому

    excelente! muchas gracias

  • @ttthr4582
    @ttthr4582 18 днів тому

    Hello, I'm very new to comfyui, i have a request if its possible for you to create a workflow and show us how to generate cartoon characters by feeding in an image of the cartoon and the controlnetpose together, to finally generate an image where the cartoon stand as per the poses in the controlnet pose image

    • @techzuhaib99
      @techzuhaib99  17 днів тому

      Creating a video is very time consuming. You can email me for the requirement

  • @murcie640
    @murcie640 2 місяці тому

    Awesome Video 🤝 Thank you bro

  • @Vinny_views
    @Vinny_views 2 місяці тому

    Subscribed. 🎉

  • @Mr.Anderson-z8t
    @Mr.Anderson-z8t 2 місяці тому +3

    How do I get the SamplerCustomAdvanced node to show the image progression as it's being generated? mine doesn't show anything until it's put out to the preview image node. Please respond.

    • @techzuhaib99
      @techzuhaib99  2 місяці тому +1

      Enable previews in the manager

  • @mattm7319
    @mattm7319 2 місяці тому

    Great video and right to the point! Thanks! I can't get Reactor to work though in Flux. The error is: local variable 'model' referenced before assignment

    • @techzuhaib99
      @techzuhaib99  2 місяці тому

      Have you used my workflow?

    • @mattm7319
      @mattm7319 2 місяці тому +1

      @@techzuhaib99 yes it has this error and just don't have time so I was using FaceID instead.

  • @Jaswant7785
    @Jaswant7785 2 місяці тому

    GGUF versions which run on low ram is out , will you make video on them ?

    • @techzuhaib99
      @techzuhaib99  Місяць тому

      Coming up, with better solutions

  • @ormeleh
    @ormeleh 2 місяці тому +1

    Amazing can you provide the workflow file please?

    • @techzuhaib99
      @techzuhaib99  2 місяці тому

      github.com/techzuhaib/WORKFLOWS/blob/main/flux_with_lora_loaders.json

  • @ysinghs
    @ysinghs Місяць тому

    If we wanted to replace let’s say the flooring in one house interior image with the flooring of another house’s image like you did with the faces here. How would we go about it?

    • @techzuhaib99
      @techzuhaib99  Місяць тому

      This method will not work for that. It will need controlnet and IPadapter, depending what exactly are you trying to do. You can refer to my latest style changer video to get the idea

  • @Irokz
    @Irokz Місяць тому

    re actor just don't work with me
    Always the problem "import failed"
    i tried everything (updating comfy, etc) but i still have the same error
    and its same for every swap nodes

  • @VintageForYou
    @VintageForYou 2 місяці тому +5

    Getting error on Node import (IMPORT FAILED) ReActor Face Swap Extension Node.

    • @techzuhaib99
      @techzuhaib99  2 місяці тому +1

      'Try fix' it in manager. Or go to cmd in reactor directory and run pip install -r requirements.txt

    • @VintageForYou
      @VintageForYou 2 місяці тому

      @@techzuhaib99 I tried to fix it in the manager but not working I have already tested and run pip install -r requirements.txt Still dot working Do you have a one-click install.

    • @Larimuss
      @Larimuss 26 днів тому

      Yeah reinstall extension. Make sure you got the faveswap models

  • @Christoberchrist
    @Christoberchrist Місяць тому

    'VAE' object has no attribute 'vae_dtype' getting this error

    • @techzuhaib99
      @techzuhaib99  Місяць тому +1

      Make sure to download the right vae for flux

  • @theazraelica3984
    @theazraelica3984 29 днів тому

    I got that error when I start your workflow:
    VAEDecode
    **Exception Type:** RuntimeError
    **Exception Message:** Given groups=1, weight of size [4, 4, 1, 1], expected input[1, 16, 128, 128] to have 4 channels, but got 16 channels instead
    Can you tell me where I have to look to find the error?

    • @techzuhaib99
      @techzuhaib99  29 днів тому

      Can you make sure you are loading the flux VAE

  • @Katyfrenam860
    @Katyfrenam860 2 місяці тому

    thanks, quick question, why all the faces looks quite blurry?

    • @techzuhaib99
      @techzuhaib99  2 місяці тому

      Can you explain the issue more

  • @EpicNarrationTales-i3j
    @EpicNarrationTales-i3j Місяць тому

    Hi i want to make a short video, with 3 consistent characters in one image, using this method is it possible? Also the video has to be 16.9 format, please let me know, ive been looking ages but nothing seems to work.

    • @techzuhaib99
      @techzuhaib99  Місяць тому

      Yea, you can use reactor faceswap on them one by one.

  • @ohsocialmio3884
    @ohsocialmio3884 11 днів тому

    I could not fix it in the manager. (Reactor face swap node) no matter what i tried. I will install via Pinokkio and if that is not gonna work i am gonna try Forge.
    Edit - And a fresh installation with Pinokkio solved the issue.

    • @techzuhaib99
      @techzuhaib99  4 дні тому

      Follow the instructions : github.com/Gourieff/comfyui-reactor-node

  • @namanshonai
    @namanshonai 2 місяці тому

    Hi, in the very first step when "Load diffudion model" is used i am not able o change unet_name it always remain "undefined" and same with weight_dtype "default" please help

    • @techzuhaib99
      @techzuhaib99  2 місяці тому

      Make sure to download the model first to the correct folder ComfUI/models/unet and hit refresh again

    • @namanshonai
      @namanshonai 2 місяці тому

      @@techzuhaib99 hi thanks alot for replying. Can you please share any video of your or any other creator

  • @Amazingphysicscourses
    @Amazingphysicscourses 2 місяці тому

    hello. I could not fix it in the manager. (Reactor face swap node). How to do this using cmd. Please explain more.

    • @techzuhaib99
      @techzuhaib99  2 місяці тому +1

      Make sure to download insightface, watch this video on how to download insightface ua-cam.com/video/vCCVxGtCyho/v-deo.html

    • @ohsocialmio3884
      @ohsocialmio3884 11 днів тому

      @@techzuhaib99 hi i have the same issue and i followed the instructions of that video i keep getting the same error.

  • @zappist751
    @zappist751 Місяць тому

    Is there a way to run a lora model this way?

    • @zappist751
      @zappist751 Місяць тому

      I meant like a lora model trained on a custom face

    • @techzuhaib99
      @techzuhaib99  Місяць тому

      You can load it the same way

    • @zappist751
      @zappist751 Місяць тому

      @@techzuhaib99 I'm not too familiar with comfy ui and image generation in general. Would I have to replace the flux dev model with the lora model or do I add another load lora module?

    • @techzuhaib99
      @techzuhaib99  Місяць тому

      @@zappist751 you have to add a lora model. What exactly are you trying to do

    • @zappist751
      @zappist751 Місяць тому

      @@techzuhaib99 I have a lora model trained on a specific celebrity.

  • @luciocasellato6115
    @luciocasellato6115 Місяць тому

    I need a tutorial for this tutorial :D

  • @CsokaErno
    @CsokaErno Місяць тому

    There are no Reactor things here :)

    • @techzuhaib99
      @techzuhaib99  Місяць тому

      ??

    • @CsokaErno
      @CsokaErno Місяць тому

      @@techzuhaib99 At 4:45 you use "Reactor Face Swap" that I can't find in ComfyUI :/