How to Inpaint in Stable Diffusion A1111, A Detailed Guide with Inpainting Techniques to level up!

Поділитися
Вставка
  • Опубліковано 28 жов 2024

КОМЕНТАРІ • 40

  • @Trinketorium
    @Trinketorium 9 місяців тому +5

    If you get the civtai extension via the a1111 extension tab you can just click update and it will scan your lira folder and add images and keywords for you

  • @maddyvoldy
    @maddyvoldy 10 місяців тому +1

    Am using sd since many months, yet I didn't know a lot of stuff explained in this video. Thanks a lot. Learnt something today. Really appreciate this knowledge share

    • @KeyboardAlchemist
      @KeyboardAlchemist  10 місяців тому

      You're welcome! I'm glad this video was helpful for you!

  • @rexs2185
    @rexs2185 11 місяців тому +1

    Really appreciate this tutorial! Reinforcing these skills in a step-by-step way has made it much easier for me to emulate in my own image generation!

  • @hyperdeloutz2444
    @hyperdeloutz2444 11 місяців тому +1

    WOW!! Never knew LORAs could be organized in such a way. Learnt so much and thank you for this inpainting tutorial.

    • @KeyboardAlchemist
      @KeyboardAlchemist  11 місяців тому +1

      You are welcome! I'm glad it was helpful for you. Cheers!

  • @fishpickles1377
    @fishpickles1377 4 місяці тому +2

    Super great tips in this video, thanks!

  • @mulda1234
    @mulda1234 10 місяців тому

    Great videos, thank you! One of the hardest things to do is to fix arms. It would be great to see a video about this.

  • @ronaldp7573
    @ronaldp7573 10 місяців тому +1

    Great video all the useful information layed out an elegant manner.

  • @Electronic_Cats
    @Electronic_Cats 5 місяців тому +1

    Very nice tutorial

  • @ISIMUSI
    @ISIMUSI 5 місяців тому +1

    Very interesting video, I didn't know this tool. Since you are open to new requests, I would like to propose one that interests me..
    I am becoming used to using Automatic1111 on the renderings I produce to add some details here and there.. being high-res images I use the Tile/blur(controlnet) + Tiled combo Diffusion&TiledVae to work directly on it and do further upscaling.
    When there are people in the images, they very often come out wrong. Is there any way to use the Inpaint module directly in high resolutions without making crops?

    • @KeyboardAlchemist
      @KeyboardAlchemist  5 місяців тому

      I have a video on Tiled Diffusion / Tiled VAE, check it out here: ua-cam.com/video/44waH3sDYOM/v-deo.html. Not sure if this video answers your question or not.

  • @yss7557
    @yss7557 9 місяців тому +1

    hey dude, you are really good youtuber and I really enjoy your videos. I have one quick question, where can I get the repository or like a file selection with all the models you are using? or if you can’t provide that, can you please share your opinion on which model would be best for generating realistic and very detailed photos, sometimes NWSF. Thanks a lot!

  • @kennethgosch5638
    @kennethgosch5638 10 місяців тому +1

    Great Video! I stumbled across this Aitana Lopez thing. I found it relatively easy to recreate pictures with the same face quality using juggernaut. But they have pictures where she is in a pool for example and this certainly doesn’t look AI generated to me. Could you explain how they are pulling this off? Greetings from Norway

    • @KeyboardAlchemist
      @KeyboardAlchemist  10 місяців тому

      Thanks for watching, I'm glad you liked the video! Without knowing the exact image(s) in question, I can't provide specifics, but I imagine they probably use a combination of checkpoint and LORAs to achieve a film grain type of look for the AI influencer photos.

  • @viper5955
    @viper5955 5 місяців тому +1

    good stuff got my sub

  • @अघोर
    @अघोर 11 місяців тому +2

    Your videos are easier to understand. I like every video you publish.

    • @KeyboardAlchemist
      @KeyboardAlchemist  11 місяців тому

      I'm glad to hear that. Appreciate your support!

  • @billyn7757
    @billyn7757 10 місяців тому

    If i may ask, is it possible to make 2 poses from controlnet, generate them separately and inpaint them in 1 image?

  • @wernerblahota6055
    @wernerblahota6055 11 місяців тому +1

    👍👍👍

  • @EDashMan
    @EDashMan 6 місяців тому +1

    Is it possible to do this on an image of myself? Please do a tutorial

    • @KeyboardAlchemist
      @KeyboardAlchemist  6 місяців тому

      Yes, definitely possible to do it with a real photo. The procedure is basically the same. I may try to work on a video like that in the future. Thanks for your suggestion.

  • @LouisGedo
    @LouisGedo 11 місяців тому +1

    👋

  • @CoconutPete
    @CoconutPete 7 місяців тому

    still confused about whole picture vs. masked only... seems like I've seen tutorials that recommend the opposite. Will need more practice lol

  • @yibaoi
    @yibaoi 10 місяців тому

    Hey nice artwork! Just want to ask that are you interested in our new ai generation product? Maybe I can send you beta version to you~

  • @CoconutPete
    @CoconutPete 7 місяців тому

    photopea remove background doesn't work with A1111 anymore

  • @nemonomen3340
    @nemonomen3340 11 місяців тому +1

    Haven’t heard of Photopea before, but people probably pronounce it that way because they think it’s a smash up of “photo” and “utopia” but that would be wrong since the spelling is off.

  • @DigitalWerkstattDE
    @DigitalWerkstattDE 11 місяців тому +1

    Hey, i practice every day. But I dont find a way to add realistic normal glasses. Its anoying, I get everything but not realistic results.

    • @cyberprompt
      @cyberprompt 11 місяців тому +3

      CivitAI has a bunch of "glasses" LORAs. If your checkpoint model can't do glasses (and many can just by saying "wearing glasses") try adding a glasses lora and inpainting with that "wearing glasses" positive prompt. Also, negative prompt "missing eye glasses". If you have any graphics skills find a photo of the glasses you want and use the technique shown here. You can also use After Detailer for eyes or face detection with the positive prompt "wearing eyeglasses" as well as the negative "not wearing eyeglasses". I can get a starter image with something like "secretary outfit with glasses." Sometimes adding it with a type of outfit or profession that stereotypically has a person who wears glasses also helps.

    • @DigitalWerkstattDE
      @DigitalWerkstattDE 11 місяців тому +1

      @@cyberprompt Thank u, very much. I didnt say, that I always work with real photos. Not often I create Ai images from scratch. Most time I inpaint in photos. This is why I hv problms with glasses. I try to inpaint glasses on real persons.

    • @KeyboardAlchemist
      @KeyboardAlchemist  11 місяців тому

      Thank you for the sub friend! I agree with cyberprompt's suggestions, it's probably easier to find a photo of a pair of glasses or a LORA for glasses, then inpaint them into your photos.