CivArchive
    Preview 1823620
    Preview 1823624
    Preview 1823623
    Preview 1823622

    DreamShaper - V∞!

    Please check out my other base models, including SDXL ones!

    Check the version description below (bottom right) for more info and add a ❤️ to receive future updates.
    Do you like what I do? Consider supporting me on Patreon 🅿️ to get exclusive tips and tutorials, or feel free to buy me a coffee
    🎟️ Commissions on Ko-Fi

    Join my Discord Server

    For LCM read the version description.

    Available on the following websites with GPU acceleration:


    Live demo available on HuggingFace (CPU is slow but free).

    New Negative Embedding for this: Bad Dream.

    Message from the author

    Hello hello, my fellow AI Art lovers. Version 8 just released. Did you like the cover with the ∞ symbol? This version holds a special meaning for me.

    DreamShaper started as a model to have an alternative to MidJourney in the open source world. I didn't like how MJ was handled back when I started and how closed it was and still is, as well as the lack of freedom it gives to users compared to SD. Look at all the tools we have now from TIs to LoRA, from ControlNet to Latent Couple. We can do anything. The purpose of DreamShaper has always been to make "a better Stable Diffusion", a model capable of doing everything on its own, to weave dreams.
    With SDXL (and, of course, DreamShaper XL 😉) just released, I think the "swiss knife" type of model is closer then ever. That model architecture is big and heavy enough to accomplish that the pretty easily. But what about all the resources built on top of SD1.5? Or all the users that don't have 10GB of vram? It might just be a bit too early to let go of DreamShaper.

    Not before one. Last. Push.

    And here it is, I hope you enjoy. And thank you for all the support you've given me in the recent months.

    PS: the primary goal is still towards art and illustrations. Being good at everything comes second.


    Suggested settings:
    - I had CLIP skip 2 on some pics, the model works with that too.
    - I have ENSD set to 31337, in case you need to reproduce some results, but it doesn't guarantee it.
    - All of them had highres.fix or img2img at higher resolution. Some even have ADetailer. Careful with that tho, as it tends to make all faces look the same.
    - I don't use "restore faces".

    For old versions:
    - Versions >4 require no LoRA for anime style. For version 3 I suggest to use one of these LoRA networks at 0.35 weight:
    -- https://civarchive.com/models/4219 (the girls with glasses or if it says wanostyle)
    -- https://huggingface.co/closertodeath/dpepmkmp/blob/main/last.pt (if it says mksk style)
    -- https://civarchive.com/models/4982/anime-screencap-style-lora (not used for any example but works great).

    LCM

    Being a distilled model it has lower quality compared to the base one. However it's MUCH faster and perfect for video and real time applications.

    Use it with 5-15 steps, ~2 cfg. IT WORKS ONLY WITH LCM SAMPLER (as of December 2023, Auto1111 requires an external plugin for it).

    Comparison with V7 LCM https://civarchive.com/posts/951513

    NOTES

    • Version 8 focuses on improving what V7 started. Might be harder to do photorealism compared to realism focused models, as it might be hard to do anime compared to anime focused models, but it can do both pretty well if you're skilled enough. Check the examples!

    • Version 7 improves lora support, NSFW and realism. If you're interested in "absolute" realism, try AbsoluteReality.

    • Version 6 adds more lora support and more style in general. It should also be better at generating directly at 1024 height (but be careful with it). 6.x are all improvements.

    • Version 5 is the best at photorealism and has noise offset.

    • Version 4 is much better with anime (can do them with no LoRA) and booru tags. IT might be harder to control if you're used to caption style, so you might still want to use version 3.31.

    • V4 is also better with eyes at lower resolutions. Overall is like a "fix" of V3 and shouldn't be too much different.

    • Results of version 3.32 "clip fix" will vary from the examples (produced on 3.31, which I personally prefer).

    • I get no money from any generative service, but you can buy me a coffee.

    • You should use 3.32 for mixing, so the clip error doesn't spread.

    • Inpainting models are only for inpaint and outpaint, not txt2img or mixing.

    Original v1 description:
    After a lot of tests I'm finally releasing my mix model. This started as a model to make good portraits that do not look like cg or photos with heavy filters, but more like actual paintings. The result is a model capable of doing portraits like I wanted, but also great backgrounds and anime-style characters. Below you can find some suggestions, including LoRA networks to make anime style images.

    I hope you'll enjoy it as much as I do.

    Official HF repository: https://huggingface.co/Lykon/DreamShaper

    Description

    NOT for training and NOT for txt2img.
    Should only be used for inpainting and outpainting.

    FAQ

    Comments (75)

    shekelbergAug 1, 2023
    CivitAI

    Is this gonna be the last 1.5 DreamShaper model? before you move permanently to SDXL?

    RizzlordAug 2, 2023
    CivitAI

    StableDiffusionXLPipeline.__init__() got an unexpected keyword argument 'safety_checker' ?

    BariiiAug 2, 2023
    CivitAI

    Is there any tip to generate better male images? All images with "male" prompt have almost identical hairstyle and face, and they all look at least 50+ years old…

    razunterAug 2, 2023· 1 reaction

    add name / race / location

    SD_AI_2025Aug 3, 2023

    PROMPT.

    If you don't say what you want or don't want, you'll get always the same.

    ChocoboyMLGAug 3, 2023
    CivitAI

    What is better, version 8 or sdxl 1.0 ?

    ktiseos_nyxAug 3, 2023

    My answer:
    Depends on what you're looking for.

    If you're still wanting 1.5, pick up 8.

    If you want SDXL - go for SDXL <3

    ChocoboyMLGAug 3, 2023

    @duskfallcrew but I mean, does SDXL model make good images according to the prompts you tell it or is it still too early and should I continue with model 8 until they improve it? Thanks for answering me and sorry for my english xd

    ktiseos_nyxAug 3, 2023

    @ChocoboyMLG No you're fine but it's the same answer, see the pics in each model and see what's your preference

    If you dont have at least 12GB of ram on your graphics card you may have trouble with sdxl based stuff. I have 5GB on mine (low I know) and i cant use sdxl with out getting out of memory errors.

    patriciagomesoo721Aug 10, 2023

    For me (my use case) SDXL is better because it has many artists styles, so I can achieve stylistic results without controlnet. For art is amazing, for realistic portraits maybe 1.5 checkpoints still wins.

    svt1701Aug 6, 2023· 3 reactions
    CivitAI

    Fuck yes! The best model! Set up a monument to this man!

    AironHeartAug 7, 2023· 6 reactions
    CivitAI

    Hi guys! I am a solo developer and I made an app: Aironheart
    At AironHeart, you can play DreamShaper without GPU for FREE!
    Try it now at https://www.aironheart.com/explore/text2image

    BurningSpyAug 7, 2023
    CivitAI

    Hey Quick question from a SD Noob here. Do I need any additional setup or settings? I've loaded the model into my sd webui and it seems to work fine. the preview shows that the image gets generated just fine, except that at the very end it totally messes up all the colors and makes it look more like a picasso painting...

    hcommand68990Aug 7, 2023· 1 reaction

    Try copying the generation data from one of the images other uses have posted. As an example, here's one of the top images:
    https://civitai.com/images/1787093?period=AllTime&periodMode=published&sort=Most+Reactions&view=categories&modelVersionId=128713&modelId=4384&postId=444009

    If you look towards the bottom of the sidebar on that image, you'll find the prompt and negative prompt* used to generate it, along with the steps, CFG scale, sampler, etc. Copy all of that and run it. Do you get a similar image?

    I did it myself, even using a different version of Dreamshaper, and got several very similar images.

    *The negative prompt in this case is just the EasyNegative embedding. I'd recommend getting it. But a good, general negative prompt you can use for everything is simply "(worst quality, low quality:1.4)". Remove the quotes, keep the parenthesis.

    BurningSpyAug 8, 2023· 1 reaction

    @hcommand68990 Thanks for the reply. I found out that there was something wierd with the loaded checkpoints, I reinstalled them and restarted and everything was fine.
    Thanks anyway!

    141397Aug 12, 2023

    Also a noob (not used SD as much yet). But when this happened to me, I had forgotten to install and use a VAE. Just FYI for another time.

    EDIT:Unless the VAE is baked in I guess.

    myk3Aug 7, 2023
    CivitAI

    Should we be using a VAE or is it baked? Thanks

    sinatraAug 9, 2023

    I think it depends on the version. if the version doesn't have a baked VAE version then it doesn't need it. If the version has backed VAE version preferably get backed VAE version.

    sinatraAug 9, 2023

    if there is a resolution problem in all the generated images, a VAE may be missing

    sinatraAug 9, 2023
    CivitAI

    I would like to leave an observation. Models 7 and 6 had a big drop in quality when compared to 5. I haven't tested the 8 very well yet. Despite this, the 7 is very interesting from the point of view that it reduces a lot in size while maintaining good quality despite being small. Maybe the versions have different purposes but the variation in quality causes a bit of confusion.

    macynkapono298Aug 9, 2023· 13 reactions
    CivitAI

    I have create a tutorial to help no GPU people to this model generate Art QRCode

    sammitra077266Aug 12, 2023
    CivitAI

    Hi,
    V8 is giving me NaN error when upscaling. I also tried using --no-half and disabled the nan check. But then it gives me only black images. I am using gtx 1650 which I think cannot use the 16 bit precision.
    Is there any way I can get it to work? Should I try the older models?

    FraxxAug 14, 2023· 10 reactions
    CivitAI

    Where can I download the VAE for the 8v?

    moshodysonAug 23, 2023· 3 reactions

    Rule of thumb for any model/checkpoint without VAE. Get the most stable generic VAE from huggingspace most people recommend vae_ft_mse_840000_f16.ckpt

    mookyAug 17, 2023
    CivitAI

    Can you use this model for commercial use?

    TheSmokingGPUAug 18, 2023
    CivitAI

    Can anyone share the secret to how to get small boobs with this model? Even when I specify (small breasts:1.5) I get pretty much "large or larger than large."

    sneky_snekAug 18, 2023· 4 reactions

    Put breasts in the negative prompts with a low weight. Some models will see "breasts" in the prompt and ignore anything else.

    FllamberAug 18, 2023· 15 reactions

    I guess it is easier to like big boobs

    BadjakAug 19, 2023· 4 reactions

    I use (flat chest:1.2)

    TheSmokingGPUAug 19, 2023

    @Badjak Thank you for your reply. I've tried that, and I still wind up with big boobs in my images.

    TheSmokingGPUAug 19, 2023

    @Fllamber I assume that's what all the models are trained on. :/ Which sucks because I could train my own model but then I'd have to learn how.

    TheSmokingGPUAug 19, 2023

    @sneky_snek I have "small breasts, medium breasts, large breasts" in the negative and "tiny breasts" in the prompt. Still, no dice. It's frustrating!

    BadjakAug 19, 2023· 1 reaction

    @mjranum150 Check my profile. I just uploaded some test data, which may be useful for you.

    TheSmokingGPUAug 19, 2023

    @Badjak Will do, thanks!

    sythraAug 21, 2023

    @mjranum150 I find that this model does very well at cute little "flat chest" characters. But it may only do it if you're referring to a younger girl. One thing that can be a factor is if you're working from a piece you like and want to change a little bit so your recycling the seed as your changing prompts it's possible that you're fighting a large breast prompt that's hidden in the seed even though you may have removed it from the prompt areas. Unfortunately there's no way to change or see what's in the seed itself

    yasirAug 21, 2023

    @sythra hi i need help can i get its documentation or code or how i can customize it example changing inference steps also it didnot generate NSFW when i try to do just shows black screen

    sythraAug 21, 2023

    @yasir black screen is a compatibility or set up issue. You referring to the model or one of my generations and how it's used in WebUi? I can show you my prompts and setting parameters etc but I don't really do NSFW so I'm not sure about how it does that it seemed to do plenty sexy for me but I've never asked it to do nudity so I wouldn't know.

    yasirAug 21, 2023

    @sythra thanks for replying i am using this model in huggingface space through gradio UI. if you have time can you provide the code

    yasirAug 21, 2023
    arnorwingAug 25, 2023· 1 reaction

    Petite sometimes works. Small breasts without brackets sometimes works. It always wants to make them (small breasts) large!

    chollman82141Aug 29, 2023· 2 reactions

    In negative prompt: "huge boobs, huge tits, huge breasts, large boobs, large tits, large breasts, busty, buxom, c-cup, d-cup, e-cup, f-cup, g-cup, h-cup"

    wilsonicsAug 30, 2023· 2 reactions

    "large breasts" in the negative prompt usually works for me. Lots of checkpoints out there seem to love 'em large.

    TheSmokingGPUSep 10, 2023· 1 reaction

    @Badjak I've been using "flat chest" and that seems to get it done! Thank you!

    fulforget85886Sep 11, 2023· 1 reaction

    (flat_chest:1.4)

    dustin4AIAug 25, 2023
    CivitAI

    I get "RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0" in Stable Diffusion.

    File Location of Error: Stable Diffusion\stable-diffusion-webui\venv\Lib\site-packages\torch\nn\functional.py

    How can I fix this?

    TillerzOct 5, 2023

    Might be a memory issue, not all of the models you use fit into your VRAM, something stays in your PCs RAM.

    ironlockadventuresSep 3, 2023
    CivitAI

    Hi everyone, I've read some comments and replies on the posts below and I understand that the images can be sold but are we able to use the images in video games? It was just unclear on the CreativeML Open RAIL-M on the limitations of commercial use.

    Thank you

    SC13NC3F1CT10NSep 7, 2023

    Technically yes, I think. However, if you want to publish said game, be warned that Steam doesn't allow any AI generated content in games.

    ironlockadventuresSep 7, 2023· 1 reaction

    @SC13NC3F1CT10N Thanks for the info!

    KriimaSep 8, 2023· 3 reactions
    CivitAI

    I'm still amazed by this model, and how it's able to produce SDXL quality level level images. Best model ever !

    maxpoetica465Sep 19, 2023

    Between this and photon nothing compares.

    vizsumitSep 11, 2023· 12 reactions
    CivitAI

    Please publish full model for v7 and v8. Full version are good for training LoRA.

    aiartmodelsSep 22, 2023· 3 reactions
    CivitAI

    Please sir, provide full parameters for text2img, img2img generation, (v4, v8 specific)

    rajendrankunduSep 30, 2023
    CivitAI

    What is "Hash AUTOV2: 879DB523C3"?

    TillerzOct 5, 2023

    If you click that hash, you can see the model has several ones. And AUTOV2 is one of them. So depending on what's being used to create the hash, the first part of that string is the name of that hashing tool.

    Indelible_SoulOct 24, 2023· 3 reactions
    CivitAI

    The inpainting model is completely ignoring my prompts, I switch back to the regular model and it follows the prompt exactly. I'm doing inpainting and just found this to be funny as hell!

    panohe27Oct 24, 2023· 9 reactions
    CivitAI

    A perfect model that can generate every type of image

    Josh54Oct 24, 2023

    You are right

    lizardbeth21908Oct 27, 2023

    Dreamshaper is what i always recommend people learn stable diffusion on its pretty forgiving , works well with most loras and is overall just enjoyable to use.

    elejamaOct 28, 2023· 2 reactions
    CivitAI

    Inpainting model has no showcase image. Please add anything, even copy the v8 standard one

    Lykon
    Author
    Dec 6, 2023

    it has multiple ones. Maybe civitai was acting weirdly?

    LiambNov 10, 2023· 7 reactions
    CivitAI

    One of my favorite models

    ellowyn101844Nov 13, 2023· 3 reactions
    CivitAI

    the hands of the generated characters are constantly sticking to different surfaces. If the character is standing, then a stone or table will be generated, and the hand will stick to it. (dreamshaper8)

    dstsckrNov 14, 2023· 4 reactions
    CivitAI

    Amazing model for mashing up vastly different LoRAs & getting what you had in mind. Thanks

    fuselayerNov 18, 2023· 5 reactions
    CivitAI

    The inpainting model is literally perfect. At least every time I've used it so far.

    neon_sin459Nov 25, 2023

    thanks to your comment I realized there was a seperate model for inpainting.

    thesolitaryowlNov 21, 2023· 11 reactions
    CivitAI

    This is my favorite model, but just curious about something, was this model trained ethically? Asking because I want to know know how to use this model personally.

    _m3phisto_Nov 22, 2023· 5 reactions

    You might want to define 'ethically', since that's a very relative term:)

    Lykon
    Author
    Dec 6, 2023· 2 reactions

    it doesn't include any opted-out artist (at least at the time of training), if that's what you mean :)

    chzbroDec 7, 2023· 9 reactions

    I've been assured that no animals were injured in the training of this model. I'm not sure about artists, but here in America we consider them nuisance mammals anyway.

    nerdygeekDec 9, 2023

    It's OK for Vegans

    efastcurexNov 23, 2023· 12 reactions
    CivitAI

    This is the model I use most for animation and I've written a tutorial on how I make Animation

    Lykon
    Author
    Dec 6, 2023

    you're gonna love the LCM version I just posted :)

    Checkpoint
    SD 1.5

    Details

    Downloads
    49,699
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/1/2023
    Updated
    5/12/2026
    Deleted
    -

    Files

    Available On (3 platforms)

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