CivArchive
    Real Humans - v1.0
    Preview 4587751
    Preview 4587748
    Preview 4587747
    Preview 4587749
    Preview 4587752
    Preview 4587750
    Preview 4587756
    Preview 4587753
    Preview 4587754
    Preview 4587759
    Preview 4587757
    Preview 4587758
    Preview 4587760
    Preview 4587763
    Preview 4587762
    Preview 4587765
    Preview 4587764
    Preview 4587766
    Preview 4590692
    Preview 4590696

    Shared originally on PublicPrompts

    strength of "1.0" is good for general use, if the faces are not real enough, increase the weight. Also closeups with extremely realistic features, need higher weight than 1

    Trigger word is "photo" and "portrait photo" forces closeups specially on higher weight


    I see huge potential for merging, but didn't test yet, it's more of an experiment that went way better than expected so I shared it

    Edit:
    - Just noticed that's it's so good for selfie prompts "selfie portrait photo of..."
    at strength 1.5
    - also noticed that I'm adding images with no prompts, feel free to ask for any missing one, I don't mind sharing

    Description

    FAQ

    Comments (22)

    7oxytronDec 15, 2023· 9 reactions
    CivitAI

    it's crazy how instantly recognizable the flickr faces hq dataset has become

    PublicPrompts
    Author
    Dec 15, 2023· 2 reactions

    It’s not directly that though, i trained on images from thispersondoesnotexist.

    PublicPrompts
    Author
    Dec 15, 2023

    but yeah the face position is definitely very recognizable

    BertramDec 16, 2023

    @PublicPrompts Which was trained on flickr faces hq (thispersondoesnotexist is StyleGAN 2 FFHQ, hence the name)

    PublicPrompts
    Author
    Dec 16, 2023· 3 reactions

    @Bertram Exactly why i used “not directly”

    KrawuzznDec 15, 2023· 10 reactions
    CivitAI

    stunning!

    KINGLIFERDec 15, 2023· 17 reactions
    CivitAI

    I am confused. What is the point of this? The whole reason many left MJ is because we wanted control of the look of our renderings. This sounds like you want the look random? Did I get something wrong?

    PublicPrompts
    Author
    Dec 15, 2023· 11 reactions

    this lora should give you MORE control not less, purpose is to make the model more capable of generating exactly the faces you want, and also not default to "standard beauty", if you want that you can still use "beautiful" in the prompt and it would work.

    also it makes super realistic skin texture

    b00merangDec 15, 2023· 7 reactions

    Not everyone here quit Midjourney, and the ones who did had different reasons for it. Apparently you are new to SD as you can't appreciate things that push the ecosystem forward.

    I think this is an amazing addition to the growing lora collection. Thanks you for the creator ❤

    1201497Dec 16, 2023

    I second what @b00merang and @PublicPrompts  said.
    @KINGLIFER This LORA complements checkpoints to bring more realism to generated pics.
    Nothing against you but try the LORA before posting a meaningless comment.

    1201497Dec 16, 2023· 1 reaction

    @PublicPrompts Super nice job btw ! Your LORA is neat :D

    KINGLIFERDec 17, 2023· 6 reactions

    @b00merang Take that nonsense to Reddit guy

    KINGLIFERDec 17, 2023· 8 reactions

    @ewolf You guys love to sound "smart" but say the most non-sequitur things... So you know for your next comment "Asking a question that cannot be answered with any sort of rational meaning. This is the textual equivalent of dividing by zero." is a meaningless question.... Do you see the "?" after the sentences... That means there is a question being made not a comment. Distinguish the differences. Shhisshhhh go.. go sit down in the back.

    StanleyPainDec 19, 2023· 2 reactions

    @KINGLIFER The only person not sounding smart right now is you, hth.

    PublicPrompts
    Author
    Dec 19, 2023· 7 reactions

    guys it's not worth the fight 🙏 KINGLIFER just asked some questions, nothing calls for personal attacks

    ThunkdeepDec 20, 2023· 1 reaction

    One day i had to learn the hard way that my fight wasn't everybody's fight and now that i know that i can sleep at night.

    CorbeMar 18, 2024· 12 reactions
    CivitAI

    Awesome if used in Adetailer (try different nationalities, names, ages)

    lucidzachary473Mar 31, 2024

    Can you recommend settings for adetailer? I can never seem to get it right haha. Like denoise strength and separate step count maybe?

    CorbeApr 2, 2024

    @lucidzachary473 Just to double check, you are putting the LORA in the adetailer prompt only right? If so, the default weight of 0.4 should be fine

    RefurrifyApr 14, 2024· 4 reactions
    CivitAI

    Curious: how many faces, repeats, epochs, network dimension? any other parameters you can share? Just trying to learn more about LoRAs, and yours is awesome!

    EphemGhastMay 1, 2024· 4 reactions
    CivitAI

    Works surprisingly well with everclear PNY

    Dani__la__7777May 20, 2025
    CivitAI

    dumb question incoming: Is this lora compatible with character embedding or will it over write the character?