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
Examples were made with the baked vae version using ft mse vae.
FAQ
Comments (55)
I'll preface by saying 1 I love your work and your Dreamshaper model, one of my favorites since I got into AI a couple months back, and also that I haven't actually used inpainting IIRC with Dreamshaper, it usually gets stuff pretty right! But I was just curious if there will be an inpainting ver for V4, or should I hang onto the old inpainting model for now (just in case)?
Of course there will be v4 inpainting!
Probably tomorrow or by the end of the week. For now you can use v3 and it will work fine as the style should be similar enough. V4 is mostly a "fix" for v3. Makes it understand prompts better, removes the need for "modelshoot style" and mostly fixes eyes at low resolution. Other than that, they're very similar as you can see from the examples.
@Lykon Thanks so much, much appreciated! Keep up the awesome work my dude.
unfortunately this model is terrible using with DreamBooth (face training)
Pretty sure it's fine, probably need different settings or regularization images. I had good success with v3.
@lman146 v4 should be even better than v3. Clip is fixed and has a way better prompt understanding.
@Lykon can you update the version on Hugginface? I can only train on Dreambooth putting the profile link itself "Lykon/DreamShaper"
@ianhmz with all the drama I didn't get enough time. I still have to upload the inpainting version and the fp16 version. I'll also update hf. Please give me a bit of time
@Lykon Sure, in your time. Thank you for all the work
@lman146 Did u use colab for train images? I'm using TheLastBen Colab.. an old colab that I'm used to use. I have success to train some models.. but this one I'm having problems on prompts.. is always realistic even when I low the training steps.
@ianhmz Generate 1000+ images using this model of male/female using random prompt script and use those as regularization. It worked for me. Before this I had the problem of getting realistic looking images.
This is such a wonderfully fantastic model. But I have no idea why creators INSIST on baking a VAE into models full well knowing that they don't play well with Loras and muck up dreambooth training.
Is there any way we'll see a Dreamshaper version without the baked-in VAE? @Lykon
gonna upload it tomorrow ;)
@Lykon YOU'RE A GOD.
@wktra ahah for something so easy? Just uploaded the no vae version you needed. I tend to give priority to the baked vae version otherwise people with less experience don't get good results and blame the model.
for civitai, why I didn't get a notification for the update?
don't want to be a karen, I just had a list of my favorite models and want all of them to be up to date all time
@justmaier
Civitai is like this.. when it works good it's great.. just don't always expect it work that way lol
@DarkAgent I don't meant to cause a ruckus, haha
it still in growing state, yes. it doing great, actually
Congratulations!. It still continues to be one of the best models so far. Thank you for the work you have taken to create and update it. It is one of my favorites since I have had access to the models.
amazing
did you agree with this?
Oh, cool! Thanks for sharing. Got no problem with it as long as they credit the original author. Correct me if I'm wrong, they're not doing it, right?
@Lykon they told me it's not dreamshaper and dalcefo. it's just "Dream" and "Dalce" that they mixed. lol. well. they use it commercial. i just think need to notice you also.
@artwork_dalcefo what about all the loras?
@Lykon i didn’t asked about that
@artwork_dalcefo since you already have an open channel with them, do you wanna do it?
@Lykon no. i don’t wana use that commercial. but what can i do? nothing without say get down.
@artwork_dalcefo oh sorry I forgot to update you. I reached out to them and they said they're gonna add credits and settings for all models and loras. They just need a bit of time. They also told me they plan to share revenues with the modelers in a spotify-like manner.
fantasy ai
you can still doewnload the model for free, use it on colab, huggingface, sinkin.ai... What's the problem? Cause you read "fantasy.ai = bad" somewhere?
1/5 ratings will be excluded if related to the fantasy.ai reddit drama, please only judge the model itself. Thank you!
thank you Ral. Sorry for the trouble. I'm thinking about leaving Fantasy.ai at this point. I'm sick of the drama.
Thanks.
Is there any way to know what this drama is about without bringing it here to the comment section?@Lykon
@Rowlinson check the latest aintrepreneur video.
@Lykon don't do it bro, it's not your fault, you deserve to get some financial support for your work. Don't mind the haters!
crap model not worth the download time or bandwidth
come on dude...
also why are you liking your own review.
It's actually a great model, despite the drama.
@SyntheticSunsets I wish people could avoid being idiots and just judge the model
You got a recommendation for us guy?
This model is amazing, I find the motivations for your comment to be suspect @harmonicdiffusion
@Saffron it was due to the fai thing and he was probably right to be upset. I've distanced myself from them.
Petty mofos need to realize that you can dislike someone or something they do but still be objective, or try to at least.
I stopped liking Kevin Spacey after finding out he's a creepy Pedo (which is far worse than any difference of opinion you could have with Lykon, obviously).... but Kevin Spacey is still a phenomenal actor (whose stuff I can't bring myself to watch anymore sadly).
You look like a petty moron if you say this is a 1 star model and your opinion has no value. It's such a good f'n work made by Lykon's hard work. The lowest I could reasonably see anyone putting on this would be 3 stars (it's a 5/5 from me). Anything beyond that and I know you're just being a petty unobjective moron.
I appreciate the message, but why make the analogy between me and a "creepy pedo" 😢
ahah
@Lykon I was hoping to be clear in that you're no way a creepy pedo or anything close to that nature of offense. It was just the first example that came to mind of someone who had done something truly truly reprehensible and even still through that I could be objective and admit how talented he was.
To be clear, I don't agree with the fantasy ai thing, but don't hold anything against you for seeing an opportunity to get paid for your work, but I love your model and objectively think it's one of the best, and I plan to continue using it (though if I had to be entirely honest, I feel my internal inclination to do so has lessened slightly).
@Balthazar99 that's why I preferred to remove myself from the drama. I prefer that plus getting pennies on patreon over drama and more money
No idea why people are being weird, went to look up the 'drama' and bscly found nothing besides a reddit post about a platform that wants to support model creators, not sure what's wrong with that?
Looks like v4 still performs as great as earlier versions, will make a proper review soon. (currently running big batch job on another model hehe)
Mostly just commenting to re-iterate that people should calm down & not give a good model bad rating because you dont like the creator or whatever the hell is going on.
well i assume the problem isn't about supporting model creators but maybe with the licensing idk but it seems it would be very difficult to know who to support because most of the models are merged models, models that weren't trained by the original person being supported. not saying thats the case with this model that's just my guess, i personally don't care its still free and a great model which you can still use to merge or generate with
Nice , but can you make a Daraz18aka lora?please
Details
Files
dreamshaper_4NoVae.safetensors
Mirrors
dreamshaper_4NoVae.safetensors
4384_dreamshaper_4NoVae.safetensors
dreamshaper_4NoVae.safetensors
47_dreamshaper_4NoVae.safetensors
DreamShaper_4NoVae.safetensors
DreamShaper_4NoVae.safetensors
dreamshaper_4NoVae.safetensors
dreamshaper_4NoVae.safetensors
DreamShaper_4NoVae.safetensors
DreamShaper_4NoVae.safetensors
DreamShaper_4NoVae.safetensors
DreamShaper_4NoVae.safetensors
DreamShaper_4NoVae.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.
