🚀Unleash Blistering Rendering Boosts & Face Masking in ComfyUI

Поделиться
HTML-код
  • Опубликовано: 19 дек 2024

Комментарии •

  • @DaVasQ
    @DaVasQ 5 месяцев назад +2

    Thx Grockster for this video and the shout out on McBoaty ;)

    • @GrocksterRox
      @GrocksterRox  5 месяцев назад +2

      It's an amazing node, amazing work!

  • @ArrowKnow
    @ArrowKnow 5 месяцев назад +1

    NIce! Can't wait to try it out!

    • @GrocksterRox
      @GrocksterRox  5 месяцев назад

      Awesome, I think you'll love it!

  • @Afr0man4peace
    @Afr0man4peace 5 месяцев назад +1

    This is awesome. 😮 This will might also work for my 3-stage workflow.. Shaving 70 percent will save so much time.

    • @GrocksterRox
      @GrocksterRox  5 месяцев назад

      Yes, it's totally awesome! Granted different samplers will take different amounts of time, but based on my tests, you'll get some benefit at the very least.

  • @yiluwididreaming6732
    @yiluwididreaming6732 3 месяца назад +1

    Nice video. The workflows look great! Can controlnet be used in your workflows to help guide the image result? If so how? BTW SUBBED!!

    • @GrocksterRox
      @GrocksterRox  3 месяца назад

      Thank you so much! For SDXL, you can definitely implement controlnet (here's an example for using depth to control posing - ruclips.net/video/o3rVsRfn2HI/видео.html). Good luck!

  • @voxyloids8723
    @voxyloids8723 5 месяцев назад +1

    Thank you for the video & what do you use for webcam face generation?

    • @GrocksterRox
      @GrocksterRox  5 месяцев назад

      Thank you so much! I've been switching off between Hedra and other internally baked workflows.

  • @97BuckeyeGuy
    @97BuckeyeGuy 5 месяцев назад +1

    Another outstanding video, sir. Thank you for the explanation on Segs and the Detailer. I feel like you could do an entire other video on how to use them. One request for future videos... Consider using the standard ComfyUI node connecting strings instead of the straight lines. The noodles may feel a bit messy, but they're easier to track than straight lines the disappear behind a node or lay on top of another line. Thanks!

    • @GrocksterRox
      @GrocksterRox  5 месяцев назад +1

      Great feedback all around, I really appreciate it! Anything in particular relating to the Segs/Detailer that you feel would be helpful to learn more about (thinking of real life cases)?

    • @97BuckeyeGuy
      @97BuckeyeGuy 5 месяцев назад

      @@GrocksterRox I'd like to just see some examples of what they can be used for. I know they're very powerful and I've seen different workflows use them, but I never fully grasped how to create them and then put them through a flow with IPAdapter and Controlnet, for example. I'd like to see an enhance/upscale workflow that uses SEG, too.

  • @WhySoBroke
    @WhySoBroke 5 месяцев назад +2

    Great videos!! Would you take on the challenge of creating photo realistic creative upscaling? Tools like SUPIR to emulate Magnific AI workflow?

    • @GrocksterRox
      @GrocksterRox  5 месяцев назад

      Definitely worth a shot, not sure if you've tried Boaty but it's a great solution in the meantime (ruclips.net/video/wDJm0qkgIOE/видео.html)

  • @StringerBell
    @StringerBell 5 месяцев назад +2

    Hey, awesome tutorial! Do you use Hedra for the talking head?

  • @bgtubber
    @bgtubber 5 месяцев назад +1

    Hm... I'm not really seeing a speedup with the new ipndm_v sampler VS dpmpp_2m in my limited tests. 🤔 At the same steps, both ipndm_v karras and dpmpp_2m karras are withing 3-4% of each other. Am I doing something wrong?

    • @GrocksterRox
      @GrocksterRox  5 месяцев назад

      Thanks for this - yeah in talking to a few folks it seems you'll get different benefits based on what your current sampler range is. Based on the quality levels I've seen with Heunpp2/Karras compared to others, I've seen at least some benefits compared to others, but you're right that it's not fool proof if you have a sampler set that's screaming fast. All that being said, it's worth doing a side by side to judge quality and speed together (e.g. I had the same speed as a Align Your Steps set, but the quality with the ipndm_v was vastly better for the same time).