LoRA Caption Workflow for ComfyUI

Поделиться
HTML-код
  • Опубликовано: 11 дек 2024

Комментарии • 54

  • @delal-ozdemir
    @delal-ozdemir 3 месяца назад +3

    u have to post even more AI / Runpod / Lora / Flux etc. because your tutorials are soooo well explainedddd. THANK YOUUUUU

  • @equilibrium964
    @equilibrium964 3 месяца назад +1

    Good job, dude. That is exactly what I was looking for! Thanks!

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад +1

      Glad you found it useful. It was based on a request in the comments of another video.

    • @equilibrium964
      @equilibrium964 3 месяца назад +1

      @@weirdwonderfulaiart I just tested it with 70 images in 8K Res, took a while, but worked perfectly.

  • @goyal_pramod
    @goyal_pramod 3 месяца назад +2

    Great workflow. Thank you!!!!

  • @vivekkarumudi
    @vivekkarumudi 2 месяца назад +1

    Thanks it worked , also is very fast.

  • @DelondialLottracus
    @DelondialLottracus 2 месяца назад +2

    Tried a bunch, feels like they were all broken. But yours worked great; actual magic. Really appreciate you sharing this. Do you have any recommendations for actual Lora trainers in comfy? Seems like most are bork'd. Now worries if you don't. Thanks a ton all the same.

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  Месяц назад

      It works for most people but it could be something to do with your current install of ComfyUI. For LoRA training should look at Fluxgym

  • @ksottam
    @ksottam 3 месяца назад

    Working great and thanks for the vid!
    Edit: Was able to get it to work on one test image and then a 'list index out of range' error or 'cannot access local variable 'image1' where it is not associated with a value' on Runpod. Tried several pods/installs, all same result

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад

      On runpod can you make sure the directory path is correctly defined - Linux format.

  • @StringerBell
    @StringerBell 3 месяца назад +2

    I'm running locally and all of the SimpleMath+ and Adjust Counter for B nodes are throwing up error (I have all the models, everything is updated)
    EDIT: IMPORTANT! - IF YOU HAVE THE SAME ERROR, MAKE SURE YOU CONVERT YOUR IMAGES TO PNG!
    EDIT 2 - Nope, the SimpleMath+ node simply doesn't work and throw off errors, no matter what I do.

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад

      Yes images need to be PNG. I will update the blog and description.
      Could be a conflict with another node! Try to setup a clean ComfyUI install and test again.

  • @kneifelkneifel
    @kneifelkneifel 3 месяца назад +1

    Amazing workflow, thanks a lot! Is it also possible to deactivate the wd14 tagger?

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад

      certainly but you need to re-route some connections to make it work. Question is why would you want to?

    • @kneifelkneifel
      @kneifelkneifel 3 месяца назад

      @@weirdwonderfulaiart thanks for your prompt response. I train some character loras for flux and it seems as wd14 tagging is not really good, so it would be nice to have the option to dis/enable it as you did with Florence 2 vs. Clip Interrogator (which is btw an awesome option to choose between!)

  • @bambaboyzstudio
    @bambaboyzstudio 3 месяца назад +1

    Thanks for this! I don't get to understand what's the purpose of this. Do you have any follow up videos?

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад +1

      Purpose is to caption your images automatically in ComfyUI which is not easy to do and then use them for LORA training. As shown in this video how to Train Flux LoRA ruclips.net/video/XEIINMF_xt4/видео.htmlsi=WYoJd5HZ19uSBkf3

    • @bambaboyzstudio
      @bambaboyzstudio 3 месяца назад

      @@weirdwonderfulaiart I see. Thanks for your response. Why I don't get to see the top bar menu with the queue batch option?

  • @blackpantherAI
    @blackpantherAI 3 месяца назад +1

    Nice work, Can i use this to caption my dataset for FLUX lora training as well? or it for stable diffusion

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад +1

      🙏 Thanks. Yes you can do that for any Lora training including Flux

    • @blackpantherAI
      @blackpantherAI 3 месяца назад

      @@weirdwonderfulaiart Great. Thanks

  • @cilginflix
    @cilginflix Месяц назад +1

    can we make our face lora with this workflow?

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  Месяц назад

      You can certainly create the captions but not train a lora with this. Checkout fluxgym for Lora training

  • @poonchecka8469
    @poonchecka8469 3 месяца назад +1

    What's the fix for this?
    PreviewImage
    'NoneType' object is not subscriptable

  • @tr0ublem4kerWZ
    @tr0ublem4kerWZ 2 месяца назад

    May I know the reason of combining Florence and WD14? thanks

  • @davidhandford1006
    @davidhandford1006 3 месяца назад

    LoRA Caption Load
    local variable 'image1' referenced before assignment

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад

      Are all the images PNG files? I didn't see this error before in my testing

    • @reijo121
      @reijo121 3 месяца назад +1

      @@weirdwonderfulaiart I have the same error. The images are jpegs.

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  2 месяца назад +1

      @@reijo121 convert them PNG. The node only works with PNG images

  • @Vinque
    @Vinque 3 месяца назад

    It works but my .png in the ui change the color like... some kind of vaporwave colorful negative, and it reads the colors that way.

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад

      Should not affect the prompt though. I didn't notice any difference in my usage which is strange.

  • @delal-ozdemir
    @delal-ozdemir 3 месяца назад

    when i do this on runpods it doesnt work ,like im running into an error when i queue it , the error comes at lora caption loader

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад

      Did the images load? Or it's the error showing before.

    • @pointy7771
      @pointy7771 3 месяца назад

      @@weirdwonderfulaiart
      I think it's the same one I'm getting:
      Error occurred when executing LoRA Caption Load:
      local variable 'image1' referenced before assignment
      File "G:\ComfyUI\ComfyUI\execution.py", line 317, in execute
      output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
      File "G:\ComfyUI\ComfyUI\execution.py", line 192, in get_output_data
      return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
      File "G:\ComfyUI\ComfyUI\execution.py", line 169, in _map_node_over_list
      process_inputs(input_dict, i)
      File "G:\ComfyUI\ComfyUI\execution.py", line 158, in process_inputs
      results.append(getattr(obj, func)(**inputs))
      File "G:\ComfyUI\ComfyUI\custom_nodes\Image-Captioning-in-ComfyUI\LoRAcaption.py", line 148, in captionload
      return text, path, image1, len(images)

    • @delal-ozdemir
      @delal-ozdemir 3 месяца назад

      @@weirdwonderfulaiart no everything was ready, then when i pressed on queue

  • @tango12341234
    @tango12341234 12 дней назад

    But dont we need the descriptions on the actual image itself and not a separate text file?

  • @blackpantherAI
    @blackpantherAI 3 месяца назад

    I am having a issue, CLIP Interrogator is throwing error, Anyone is getting the same?

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад +1

      What's the error? Did you update all components in ComfyUI, if not then try that?

    • @blackpantherAI
      @blackpantherAI 3 месяца назад

      @@weirdwonderfulaiart Yes i did updated all things, there was a long error in the console. but when i removed the CLIP Interrogator everything is working fine

  • @sandeepm809
    @sandeepm809 3 месяца назад

    Does not work it loads all the files and then shows error on maths node

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад

      make sure everything is up to date in your ComfyUI. this kind of thing can happen if things are outdated.

  • @loktar00
    @loktar00 3 месяца назад

    Nothing but errors unfortunately, nice explanation but moving onto another solution.

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  3 месяца назад

      Probably ComfyUI is out of date if you update it would help.

  • @jbeen8three
    @jbeen8three 2 месяца назад

    i don't think the wd14 tagger is working

    • @weirdwonderfulaiart
      @weirdwonderfulaiart  2 месяца назад

      Pls Update your ComfyUI and all custom nodes

    • @jbeen8three
      @jbeen8three 2 месяца назад

      @@weirdwonderfulaiart I did this several times

  • @sethmunich4062
    @sethmunich4062 Месяц назад

    error cannot access local variable 'image1' where it is not associated with a value