CivArchive
    Preview 17251514
    Preview 17251504
    Preview 17251509
    Preview 7661577
    Preview 7661579
    Preview 7661580
    Preview 7661581

    Styles are trained on synthetic data and do not copy the styles of artists!

    Photo 2 - is trained on real data (Dataset from Unsplash)

    [M] Merged LoRAs

    Recommended LoRA Strength (Weight): 0.8-1

    Preview generated without using ADetailer & Hires. fix

    Description

    Repeats: 20

    Epoch: 10

    Image Count: 24

    Network Rank: 32

    Network Alpha: 16

    Batch size: 2

    Clip skip: 1

    Training steps: 2400

    Resolution: 768x1024

    Train Resolution: 1024x1024

    Model: ponyDiffusionV6XL_v6

    Captioning: WD14 Captioning + Edit (SmilingWolf/wd-v1-4-moat-tagger-v2)

    Trigger Words: gothic art, oil painting, traditional media, realistic,

    LoRA Weight: [0.7-1]

    FAQ

    Comments (23)

    3703024Mar 9, 2024· 3 reactions
    CivitAI

    Your Lora's are great! Thank you for making these

    obuyb404
    Author
    Mar 9, 2024

    I'm glad you like My Styles and thanks for the Buzz!

    Inevitable_Kick1922722Mar 10, 2024
    CivitAI

    What are you setting for the instance and class prompts when training these? My experiments have been failures so far :(

    obuyb404
    Author
    Mar 10, 2024

    I don't do anything special, my settings are just a preset with slight modifications, I don't even remember what preset it was

    I can even say I have the laziest approach to LoRA training.

    Captioning and Trigger Words in .txt files nothing else is there

    All I have are folders with pictures and text files

    HornyReiMar 15, 2024· 8 reactions
    CivitAI

    Something broken. Only gothic is available. Anime night is missing for example

    a1b2c3mklMar 16, 2024

    Most loras are listed as XL 1.0 due to a bug some weeks ago, now civit ai made it so that you can no longer use xl 1.0 loras with pony xl. If you download it, it works just fine for pony, but until civit allows xl 1.0 being used for pony, or the author reverts them all back to be listed as pony, you wont be able to use on site generation.
    Or so i guess xd

    DoookieMar 16, 2024

    @b1te It’s so funny to me how many issues this site has. I totally understand but still

    thequackzMar 16, 2024
    CivitAI

    Please fix. Gothic is nice but I really liked using smooth anime.

    obuyb404
    Author
    Mar 16, 2024· 5 reactions
    CivitAI

    Changed the Base Model back to Pony.

    Everything should work fine!

    mikeshmikesmith656Mar 16, 2024

    Other styles can't be chosen from the generator, I think civit.ai team made some changes and broke something, hopefully this will be resolved.
    Update: But you can manually add other styles from their page by pressing "create."
    Also scale is limited to 0.7 to 1.0 for some reason.

    obuyb404
    Author
    Mar 16, 2024

    @mikeshmikesmith656 Strength is set when the model is published, I can change this to Min Strength 0.5 if you need it

    mikeshmikesmith656Mar 16, 2024

    @prgfrg23 It was fine when it was available in scale from -1.0. to 2.0 I think, always nice to have a bigger field to experiment even if you get weird results.

    obuyb404
    Author
    Mar 16, 2024

    @mikeshmikesmith656 With a value of -1.0 or 2.0, all you get is distortion and nothing else, There are no fields for experiments

    All I can offer is Min Strength 0.5 or remove Min Strength, Many models do not even have Trigger Words, so the strength of the style is not so strong and if the Strength is less than 0.5, then the style will practically not be visible

    a1b2c3mklMar 16, 2024

    @prgfrg23 i have no idea how this works, but i was using negatives and 1.3+ to get some cool effects without any distortions, examples:
    rainbow on negatives (-0.6) gives a more dull, realistic color style with everything else remaining anime style
    smooth anime (1.3) and (1.5) works wonderfully fine
    concept art twilight (1.5) or even (1.7) can generate amazing starry night skies images, or very cool edgy long coats, foggy backgrounds and even monsters
    photo on negatives (-0.5) does something weird to color, paired with high smooth anime can give a cool effect removing some contrasts and making colors more uniform

    mikeshmikesmith656Mar 16, 2024

    @prgfrg23 sometimes I want to add just a little bit, like 0.2-0.3 for line art for example. So having full range is better.

    obuyb404
    Author
    Mar 16, 2024

    I published a post to make it clear

    -1.0. -0.5 1.0 2.0

    https://civitai.com/posts/1741495

    dfsj98jfndasjfvjm604Mar 16, 2024
    CivitAI

    Why can we only set strength as low at 0.7? i like mixing and matching them on lower strength to make hybrid styles at like 60 20 and 40 40

    obuyb404
    Author
    Mar 16, 2024· 2 reactions

    Because I set Strength 0.7 when publishing models, now I will change the value to Min Strength 0.0 Max Strength 2.0

    obuyb404
    Author
    Mar 16, 2024· 1 reaction

    List of mixed models

    Smooth Anime - Smooth + Anime 50/50

    Smooth Anime 2 - Smooth 2 + Anime 2 50/50

    Smooth Anime Night - Smooth Night + Cold Night 50/50

    Oil Gothic - Oil Painting + Gothic Art 50/50

    Concept Art Twilight - Concept Art + Concept Art Dark (Unpublished) + Concept Art Negative (Unpublished) 35/35/30

    Cold Oil Gothic - Cold Oil + Gothic Art Sharp (Unpublished) 50/50

    brazenlyMar 20, 2024

    @prgfrg23 could you update previous models to reflect this? I also like to blend models!

    obuyb404
    Author
    Mar 20, 2024

    @brazenly I'm not quite sure what you mean by update?

    3678994Apr 2, 2024

    @prgfrg23 You said initially you set strength at 0.7 when publishing, then said you will change the value of min strength to 0.0 and max to 2.0. What the user meant was....have you applied/updated all the styles with this change?

    obuyb404
    Author
    Apr 2, 2024

    @Eshemizu All styles here have a Min Strength 0.0 and Max Strength of 2.0