How I work when designing ComfyUI workflows

Поділитися
Вставка
  • Опубліковано 12 січ 2025

КОМЕНТАРІ • 8

  • @antichitati.si.trandafiri
    @antichitati.si.trandafiri Місяць тому

    Very interesting insight! I remember when I first started learning Photoshop, around high school times in 2004, that I found it hard at the beginning. It was all new compared to my Fine Art training with a brush, but as you have said, the learning process was worth the effort and it enabled me to explore mediums and create things at another level.

  • @johntnguyen1976
    @johntnguyen1976 2 місяці тому

    100% home run video, as always. Thank you! 🙏🏽

  • @CoreyJohnson193
    @CoreyJohnson193 2 місяці тому +2

    Exactly what I need! Im about make a series of intense character creation workflows and need this kind of stuff.
    Also, my brain thinks in nodes... I dont know why, but I found ComfyUI easier to use than A1111 or Forge. Invoke is awesome; it has nodes and a very user-friendly UI, but is limited. I think ease is second to ability. Sadly, the AI community has spwned those looking for the easy way rather than the most efficient and proficient way to use these tools.

  • @8561
    @8561 2 місяці тому +1

    Great high level breakdown! I would agree communication is vital as many clients do not understand the tools to the extent to understand common truths such as "it's basically impossible to design a workflow that works 100% of the time".
    Curious if you can elaborate on what you mean by "set up automation" at 20:10. Do you mean engineers building dockers and API's to host the workflow and wrap a UI for ease of use?

    • @risunobushi_ai
      @risunobushi_ai  2 місяці тому +2

      Setting up automation may be as simple as setting up a batch loader pipeline and targeting a big dataset of input images, or as complex as setting up scripts to automate Lora selections depending on input images, variables selections depending on testing cases, etc.
      Not necessarily things that need a frontend, but things that mostly need to be properly tested and vetted on a large dataset, which if done manually would take a ton of time

  • @MicheleBrugiolo
    @MicheleBrugiolo 2 місяці тому +3

    Ciao Andrea molto interessante grazie! Sarebbe bello vedere anche un esempio dove tu parti da una richiesta anche inventata e ci fai vedere come ragioni e soprattutto come metti in pratica le tue idee 😊

    • @risunobushi_ai
      @risunobushi_ai  2 місяці тому +1

      I used to be doing exactly this on livestreams, but it’s been a while since I’ve had the time to do one!

  • @2.718_
    @2.718_ Місяць тому +1

    Si vous voulez vraiment apprendre à mieux diriger les ia génératives de ce genre, apprenez ComfyUi
    Après c’est assez complexe 👀