CivArchive
    Preview 707744
    Preview 707747
    Preview 707794
    Preview 707743
    Preview 708097
    Preview 707746
    Preview 707748
    Preview 707749
    Preview 707918
    Preview 707742
    Preview 728864
    Preview 728800

    NeverEnding Dream (a.k.a. NED)

    This is a dream that you will never want to wake up from

    Add a ❤️ to receive future updates. This took much time and effort, please be supportive 🫂
    Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee

    V1.22 update: thanks to Bokus for some of the preview images and the tests.

    Available on the following websites with GPU acceleration:

    It has been a while since I made a model. I've been making LoRA's non stop for the last month, but I felt like I needed something. A big problem with DreamShaper is that it wasn't much compatibles with the LoRA's I was making. So I wanted to create a model that was "artistic" like DreamShaper but also able to use my LoRA's and booru tags for generating images. Plus I wanted this to be able to make good anime stuff out of the box, without needing an additional LoRA.

    After a lot of tests and also fine tuning (plus LoRA merges) here is my newest model.

    It came out a bit more realistic then I wanted, but I won't complain.

    What is this model great at?

    • Generating cosplay images

    • Generating anime pictures

    • Work accurately with character LoRA

    • Generating good looking people

    • Generating realistic animals

    • Generating images using booru-like tags

    What is this model not too good at?

    • Generating pictures from complex sentences

    • Making fantasy/sci-fi paintings

    So this basically complements DreamShaper, and it also doesn't include Dreamlike in the mix.

    I hope you'll enjoy it!

    Some tips:

    • Use CLIP skip 2.

    • Select "auto" as vae if you're using the baked vae version.

    • Use highres fix or img2img to upscale the images after you get a preview. See the examples generation data for settings suggestions, I've tested various techniques.

    • If you're making images where the subject is far, remember you can inpaint eyes and faces selecting "only masked".

    • Stable Diffusion is great, have fun with it!

    Description

    Now with Epi Noise Offset

    FAQ

    Comments (30)

    MoreAIMay 6, 2023
    CivitAI

    No baked VAE possible, kindly? I've had anecdotal experience noticing models with such baked in, sometimes cause less than stellar generations when using a LORA and such. Also, what's the difference between the last versions form a few months back, to this new one?

    OlbanetsMay 6, 2023· 2 reactions
    CivitAI

    v1.22 has become more realistic :-(

    tsomeMay 6, 2023

    i mean realistic is good, but i never saw NED as a realistic model.

    The style of NED was everything for me but now i feel like the style is fading away with the realism part.

    Lykon
    Author
    May 6, 2023· 1 reaction

    @tsome you're not prevented from using old NED if you prefer it ;)
    I didn't aim specifically for realism, just for more quality with lighting and such. However, I don't find it realistic at all if you just remove "photorealistic" and related keywords from prompts.

    1201996May 10, 2023· 1 reaction

    this is good, 99% of models on civit are anime of underage girls that all look the same, use one of them. I personally welcome NED becoming more realistic

    PotatCatMay 12, 2023

    The dream is ending ༼ つ ◕_◕ ༽つ

    Lykon
    Author
    May 12, 2023· 1 reaction

    @Potatovision ??

    PotatCatMay 12, 2023

    @Lykon (~ ̄▽ ̄)~

    ritcher1May 6, 2023· 1 reaction
    CivitAI

    If possible, can you add the smaller 2gb pruned-fp16 version ?

    OlbanetsMay 6, 2023

    It is already fp16.

    Lykon
    Author
    May 6, 2023· 2 reactions

    I'll upload pruned-fp16 no vae soon (it's the one that should be used for training) and also the inpainting version

    ritcher1May 7, 2023

    @Lykon Thanks !

    PotatCatMay 6, 2023
    CivitAI

    Just to let you know the ko-fi page link, links to explore page :)

    Lykon
    Author
    May 6, 2023

    I know, it seems they suspended my account without notice. Just use the patreon page for now.
    I've already sent them a support ticket asking why

    PotatCatMay 6, 2023

    @Lykon 😥

    nigirimeshiMay 7, 2023
    CivitAI

    Hi Lykon. Thanks for the nice model.

    It seems that the CLIP key '41' named embeddings.position_ids in v1.22 is corrupted. I have no idea how to create the model, I just detected it with a tool.

    Please check if you like.

    Lykon
    Author
    May 7, 2023· 1 reaction

    It might be because of that corruption that the model works so well. I've done clip fix in the past and it made things worse.

    WZ_GenMay 8, 2023· 6 reactions
    CivitAI

    In 1.22 all girls look wayyyy too Korean/Asian, even with Negative Prompts, compared to the previous version.

    EDIT: Looks like I must have been doing something wrong before. Seem to work fine for non-Asian subejcts now.

    Lykon
    Author
    May 8, 2023

    they should look less asian. Strange.

    Lykon
    Author
    May 8, 2023

    Another user already tested it: https://civitai.com/reviews/92976
    But almost all the examples I posted have no modifiers and almost none of them looks Asian to me. Uhm.
    Are you using EasyNegative? Because that's trained on Asian stuff.

    Lykon
    Author
    May 8, 2023

    Adding a couple of examples of black and caucasian girls. Other people have posted Indian and Middle Eastern. Thanks.

    WZ_GenMay 8, 2023

    @Lykon Maybe I'm just doing something wrong. I don't use EasyNegative, typically not more than something like "ugly. deformed. Asian. closeup. text. blurry, 3d model". I'll take a look at some other example and see what I'm doing wrong. Thanks for the reply!

    Lykon
    Author
    May 8, 2023· 1 reaction

    Keep in mind the model is based on basilmix (because of booru tags compatibility), so it will always be a bit biased towards Asian girls. But every model is biased and every model is based on base SD, so you can always "tap" into SD1.5 concepts.

    WZ_GenMay 8, 2023

    @Lykon Those examples appear to be from the previous version, not 1.22 though

    Lykon
    Author
    May 8, 2023

    @Airashi_Gen Model hash: ecefb796ff, Model: NeverEndingDream_1.22_BakedVae_fp16

    Lykon
    Author
    May 8, 2023

    @Airashi_Gen it seems to be 1.22 according to my generation data and to civitai https://imgur.com/pdwgMz4.png

    WZ_GenMay 8, 2023· 1 reaction

    @Lykon Strange, civit must have forwarded me to the wrong picture, when I first clicked them they were absolutely showing ones with the older version. Now it shows the 1.22 ones, weird. Anyway, I think it's working better for me now if I use a bit more Asian-type negatives.

    Katsura9000May 9, 2023· 1 reaction

    They look less asian compared to previous models, like I have to go heavier with asian prompts or asian loras, not an issue at all.

    Checkpoint
    SD 1.5

    Details

    Downloads
    93,514
    Platform
    CivitAI
    Platform Status
    Available
    Created
    5/6/2023
    Updated
    5/11/2026
    Deleted
    -

    Files

    neverendingDreamNED_v122BakedVae.safetensors

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.