• aubertlone@lemmy.worldEnglish
    6·
    1 year ago

    Keep up the good work

    I’m finally upgrading to a nice rig next month, so image generations will take 1-2 seconds and not 26 like currently…

      • aubertlone@lemmy.worldEnglish
        1·
        1 year ago

        It’s on its way.

        Can I brag for just a second?

        I got a new job in August and been saving. Getting Lenovo legion 9i pro with the 4090 upgrade.

        It’s in the mail already.

        I’ll share that on my old rig, images generated in about TWENTY-SIX seconds…

        It was a little painful to experiment with that lag time. Thanks again for sharing knowledge!!

    • BlaTRUS@lemmynsfw.comEnglish
      1·
      1 year ago

      So far this is pretty much the only thing I’ve wanted to mess with that my AMD GPU is keeping me from lol.

      I think it’s possible to run stable diffusion on AMD cards, it’s just more complicated and I haven’t had the drive to bother trying to set it up.

  • nudermeisters@lemmynsfw.comEnglish
    2·
    1 year ago

    What’s your process here? Are you inpainting the background with the LoRA active, or is the LoRA active alongside other things, and the subject and background and generated at the same time?

    I ask because I was trying to use the same LoRA to make a wona standing in an falling apart doorway and it really just was not wanting to add a person, or wmif it did add a person, it was like face and upper bust level subject zoom only. I was using the zrpgartists checkpoint so I don’t think it was anything with that.

    Maybe controlnet input images?

    • Cavendish@lemmynsfw.comOPEnglish
      4·
      1 year ago

      No controlnet or inpainting. Everything was generated in one go with a single prompt. I’ll sometimes use regional prompts to set zones for head and torso (usually top 40% is where the head goes, bottom 60% for torso/outfit). But even when I have regional prompting turned off, it will still generate a 3/4 / cowboy shot.

      I assume you pulled the prompt out of one of my images? If not, you can feed them into pngchunk.com. Here’s the general format I use with regional prompting:

      *scene setting stuff*
ADDCOMM
*head / hair description*
ADDROW
*torso/body/pose*

      The loras that are in the top (common) section are weighted pretty low, 0.2 - 0.3, because they get repeated/multiplied in each of the two regional rows. So I think at the end they’re effectively around 0.6 - 0.8.

      prompt example 
      photo of a young 21yo (Barbadian Barbados dark skin:1.2) woman confident pose, arms folded behind back, poised and assured outside (place cav_rdrguarma:1.1),
(Photograph with film grain, 8K, RAW DSLR photo, f1.2, shallow depth of field, 85mm lens),
masterwork, best quality, soft shadow
 (soft light, color grading:0.4)

ADDCOMM

sunset beach with ocean and mountains and cliff ruin in the background ,
(amethyst with violet undertones hair color in a curly layers style:1.2),
 perfect eyes, perfect skin, detailed skin

ADDROW

choker ,
(pea green whimsical unicorn print bikini set:1.1) (topless:1.3) cameltoe (undressing, panty pull:1.4) 
(flat breast, normal_nipples :1.4),
(tan lines, beauty marks:0.6)
(SkinHairDetail:0.8)

      It may be that you’re not describing the clothing / body enough? My outfit prompts are pretty detailed, so I think that goes a long way for Stable Diffusion to determine how to frame things.

      • nudermeisters@lemmynsfw.comEnglish
        2·
        1 year ago

        i finally dropped one of your images into A111’s PNG info tab and holy shit man hahaha its so much; I think my 3060Ti will have a seizure if i even try that on 512

      • nudermeisters@lemmynsfw.comEnglish
        1·
        1 year ago

        Nah I haven’t reverse engineered your images in anyway, but I did originally download the LoRA based on your previous post! I’ve yet to mess with regional prompting at all, I’ll have to look into it.

        I do a lot more Img2Img than Txt2Img

        • Cavendish@lemmynsfw.comOPEnglish
          4·
          1 year ago

          I hope you’re not saying “reverse engineer” like it’s a negative or shady practice. I freely share all of my prompts to help people see what’s working for me, and I like to explore what’s working for everyone else. I’ve had good success with simpler prompts too, like the one for this parrot: https://civitai.com/images/3050333.

          • nudermeisters@lemmynsfw.comEnglish
            2·
            1 year ago

            Nah it was a verbose way to communicate, not throwing shade at ALL. If someone is on board with sharing PNGs woth Metadata, I wouldn’t consider it rude on the slightest to use said metadata.