CivArchive
    Anime Pastel Dream - Soft - baked vae
    NSFW
    Preview 316161
    Preview 316170
    Preview 316150
    Preview 316169
    Preview 316168
    Preview 316167
    Preview 316166
    Preview 316165
    Preview 316164
    Preview 316163
    Preview 316162
    Preview 316160
    Preview 316159
    Preview 316158
    Preview 316154
    Preview 316153
    Preview 316152
    Preview 316151
    Preview 1695213
    Preview 1695212

    Anime Pastel Dream (Soft and Hard)

    Add a ❤️ to receive future updates.
    Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee

    Available on the following websites with GPU acceleration:

    This is my attempt at making a good anime pastel model. I love the end result and I hope you enjoy it. This should also work great with most anime models.

    I made 2 versions:

    • Hard has a stronger pastel style and gets a lot more creative.

    • Soft has a subtler pastel style and is much easier to control.

    Consider supporting model creators of the models you like!

    Suggested settings:

    • CLIP skip 2

    • I had ENSD: 31337 but it's optional

    • NO restore faces

    • Highres fix with either a general upscaler and low denoise or Latent with high denoise (see examples)

    • Be sure to use Auto as vae for baked vae versions and a good vae for the no vae ones. Avoid anythingv3 vae as it makes everything grey.

    • Steps and upscale denoise depend on your samplers and upscaler. See the examples.

    Description

    FAQ

    Comments (23)

    ktiseos_nyxMar 23, 2023· 1 reaction
    CivitAI

    OH MY GOD x.x you've released a monster that my addiction to merging models is going to ... drools - I think i have a model i merged with NED btw, this is ok right? XD As long as it's not entirely poor and crappy XD (Cause NED is perfection)

    Lykon
    Author
    Mar 23, 2023· 1 reaction

    I'm not super sure about the results of a merge between them. I'm also working on the next version of NED, so stay tuned for that.

    ktiseos_nyxMar 24, 2023

    @Lykon Yea I've not publically put it in anything yet i was toying with it, i'm in a passe on how my trained models are still turning out bad but i forgot that half the time i don't tag the images XD

    AlbedowoMar 23, 2023· 7 reactions
    CivitAI

    Results look cool. Will be waiting for a prunned version.

    Lykon
    Author
    Mar 23, 2023

    sorry, gonna post it tomorrow, I promise. I'm super tired now, I need some sleep

    BrennokMar 17, 2024

    @Lykon Did you ever get around to making pruned versions? I didn't see them posted only the full models.

    wongchihonpacowo5960May 7, 2023
    CivitAI

    Which vae should i use

    Lykon
    Author
    May 7, 2023

    both have baked in vae, so you should go with Auto

    @Lykon thx

    1FunnMay 19, 2023
    CivitAI

    Idk why, but the image i generated always have bad eyes. any tips ?
    Btw im using Easy Diffusion

    Lykon
    Author
    May 19, 2023

    what are you using as prompt and resolution? Are you overwriting the vae? Are you generating full body images without highres.fix and inpainting?

    1FunnMay 19, 2023

    my prompt : 1girl, masterpiece, ultra-detailed, perfect lighting, standing, smile, (school uniform:1.1 ), long hair, small breasts, (wide hips:1.2 ), blue miniskirt, hands at sides, ((navel:1 )), blush, parted lips, outside, arms behind back, looking at viewer, blue eyes, ultra-detailed background, ultra-detailed clothing, ultra-detailed hair, tokyo background.
    negative prompt : Deformed, bad anatomy, poorly drawn face, swimsuit, extra limb, ((undetailed clothing:1.3 )), (((poorly drawn hands:1.5 ))), missing limb, floating limbs, disconnected limbs, malformed hands, out of focus, long neck, long body, (((((mutated hands and fingers:1.7 ))))), out of frame, (worst quality, low quality:1.4 ), monochrome, zombie, interlocked fingers, ((((bad eyes:1.8 )))),
    reso 512x512 upscale 2x.
    Default vae at easy diffusion (vae-ft-mse-840000-ema-pruned)
    There is no highres.fix and inpainting settings in easy diffusion.

    Im still new at this haha

    Lykon
    Author
    May 19, 2023· 2 reactions

    @1Funn it seems you're making a full body image at 512x512. You need to use highres fix or inpaint masked-only at at least 300x300 masking only the face. Or you can try generate a portrait (so no mention of skirt, legs, shoes, etc, and instead prompt for "upper body") at 512x768. This will get more detailed but it will always improve with highres.fix.

    1FunnMay 19, 2023

    @Lykon okay, i will try it

    Arc_WizardMay 21, 2023

    Try to: increase pixels, Increase the proportion of characters in the screen, Appropriate increase steps

    1FunnMay 23, 2023

    @Lykon Its work perfectly on stable diffusion webui. I think its the easy diffusion problem, because there is no highresfix in it.

    Lykon
    Author
    May 23, 2023· 2 reactions

    @1Funn you can still use an upscaler manually + img2img. It's the same thing as highres fix

    peruio2_123747Aug 25, 2023· 3 reactions
    CivitAI

    error bro

    gojosaturoApr 23, 2024· 1 reaction
    CivitAI

    Hello!, Is the “Anime Pastel Dream” model suitable for training LoRA anime character?

    NourdalApr 27, 2024
    CivitAI

    A very good anime model, however, like the rest of the Lycon models.

    jamezJun 27, 2024
    CivitAI

    What do you mean by no restore face?

    NyriniJun 29, 2024
    CivitAI

    I can't seem to use it

    eternalchimeDec 28, 2024
    CivitAI

    It lives up to its name.