AbsoluteReality
That feeling after you wake up from a dream
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 ☕
For LCM read the version description.
Additional ecamples for V1.6: https://civarchive.com/posts/353121
Additional examples for V1: https://civarchive.com/posts/259634
Amazing gallery by qf22: https://civarchive.com/posts/260939
Quick face alteration examples using celebrity names and mixing them: https://civarchive.com/posts/274268
Available on sinkin.ai, Mage.space and many other services
Suggestions
Use between 4.5 and 10 CFG Scale and between 25 and 30 Steps with DPM++ SDE Karras. Worse samplers might need more steps.
To reproduce my results you MIGHT have to change these settings:
Set "Do not make DPM++ SDE deterministic across different batch sizes." (mostly for v1 examples)
Set the ETA Noise Seed Delta (ENSD) to 31337
Set CLIP Skip to 2
DISABLE face restore. It's terrible, never use it
Use simple prompts. Complex prompts might make less realistic pictures because of CLIP bleeding. More complex prompts does not mean better results. Keep it simple.
Use ADetailer to enhance faces. Basically every solo portrait I made uses it. You can get my settings by clicking on "copy generation data". I suggest you use denoising under 0.3 to avoid getting always the same face.
Use BadDream and UnrealisticDream negative embeddings (
BadDream, (UnrealisticDream:1.2)). Add weight to UnrealisticDream between 1.2 and 1.5. Do not use FastNegative or EasyNegative if you aim at realism. However, they're good for artworks.Use Highres.fix with the following settings:
Denoising strength: 0.45, Hires steps: 20, Hires upscaler: 8x_NMKD-Superscale_150000_Gand as much upscale as you can (my gpu only handles up to x1.8 at 512x768 base resolution, but you can go higher). If you don't have8x_NMKD-Superscale_150000_Gyou can probably use another GAN, but it should be easy to find on Google. You can also try Latent with a denoise higher than 0.6, but the result will be harder to control.Try to condition faces by prompting for eye colors, hairstyles, hair color, ethnicity and so on. Even celebrity names do work. This model is pretty good at not making a single face if you play with the context.
If the pic is too clean, try to add some ISO noise. Even as a post processing with external tools it will trick the brain enough to make you think "damn, this is a real photo.
If you feel the grounded nature of this model is limiting your imagination, try generating on DS6 and then do img2img with this one to bump up the realism.
Pruned vs full comparison (not highres fixed)
Brief story of this model
While working on DreamShaper 6 I made many other models and a crap ton of tests. Some of them were strange mixes with photorealistic models I had previously made, plus new ones and some ISO noise LoRA I made. When I tested the various release candidates of DS6 for photography prompts, this came out on top. While mostly on the losing side with regards to range, flexibility and LoRA compatibility compared to what became DS6, I noticed this was pretty good at recreating photos with very simple and minimalistic prompts. So, why not, I gave it some love and kept working on it, just in time for its initial release.
Difference with DreamShaper
Long story short, DS aims at art, this aims at realism. They might overlap a bit, but they have different objectives and different things they're good at. DreamShaper is total freedom and can basically do everything at a high level. This does about one thing and does it extremely well. This still uses DreamShaper as base, so it's capable of doing art to a lesser degree.
Description
LCM version. Use it with 5-15 steps, ~2 cfg. IT WORKS ONLY WITH LCM SAMPLER (to get it on Auto1111 you currently need a plugin).
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.
FAQ
Comments (67)
Love me some AbsoluteReality! Excited bout the refreshment, LCM is the new SD :=}
Thank you lots for LCM model. I am comfyui user, Kindly tell how and where comfyui user use this "Set the ETA Noise Seed Delta (ENSD) to 31337" features.
ENSD isn't as important these days but i guess it may matter for LCM i dunno where on comfy to set it tho
Yeah you don't need it unless you want to reproduce my exact results (for which you might also need my same gpu)
This is the best image generation and inpainting model in all of Civitai. Congratulations. If I didn't live in a third world country, I would donate money to you.
How is this better than epicrealism?>
You live in United state of america? Sorry for you :)
cool
I type in a handsome 1guy, muslim and I got a guy with mustache wearing hijab. O___o
Maybe try putting a race instead of a religion like 'arab' or 'Pakistani' it might be what ur looking for ok cya laterrrr
dern, i've been trying all week to get a guy with a mustache and hijab
@thefoodmage well technically we belong to the human "race". "Arab" is an ethnicity and "Pakistani" is a nationality, so none of them are "races". You're on the right path tho 💪
How do I get more face variation with this model
i guess adding random people name inside a bracket ((),[]) will help
Loved this model so so much, will you consider to come out with a SDXL version of this?
i'd love to see an XL Version too. nevertheless this is a really great model. Was using Dreamshaper before, but now i'm almost always using Absolute Reality
@nxs2k Same here, btw I noticed that the DreamShaperXL is also capable to produce realistic photo, just the quality isn't as good as AbsoluteReality.
Can this be used as a refiner?
Any Idea why I get entirely different results although I use the png info feature? Here are the images I get in comparison
an extended dataset compared to dream sharpen?
<absolutereality_lcm> if we set the cfg to 2 the dark images appear and if we set it to 1 they appear lighter but the women appear younger.
A problem when publishing a post.
Forbidden names: Maddi Abaroa, Susanita, Melika, Kaitlynn Boothman, Jenna Ortega, Greta Thunberg.
For LCM model, where can I get the LCM test sampler. Automatic1111 doesn't have that one
The easiest way to put "LCM" sampler to sampling methods on Automatic1111 by installing AnimateDiffusion extension https://github.com/continue-revolution/sd-webui-animatediff I'm not using Automatic1111 for dailies, but i've tried and it worked.
you can add any sampler or scheduler you want to A1111. It is open source code.
I really like this model for color and ease of manipulating. I kinda wish it would do hands a little better though.
What VAE should be used with this checkpoint?
I have the feeling it has mse-840000 baked in.
Is there a chance to see this great model as an SDXL one?
With such a diverse model how do you accommodate different categories (cars, people, etc)into your training or dataset preparation?
1. Do you separate categories into different folders or do you put all the images into one dataset? And how important is captioning?
2. What's the secret to training hands well?
Although the realism of this model is not ABSOLUTE..., but it is still ABSOLUTELY excellent, just like other Lycon models!
Lycon is surely a synth pretending to be a mere mortal like us. Either way I download anything he publishes.
Works well, creates great results.
Will you upload the full non-pruned version? (1.8.1)
Can you please tell me what you mean by "add some iso noise"?
means you add a layer of fake film grain texture (as in grainy old photographs or films which are the result of low iso film) to hide smooth computer generated look
cute model!
The model doesn't follow the prompt to a T, but it does have a lot of options. The photorealism is on point, for sure. 8/10
has been trained on illegal data, capable of generate accurate illegal content whitout an explicit prompt, lot of people use that for bad thing, do better
your mind is stuck, a knife can save lives or take or even create art, depends on user, the knife is just a tool.
@Overloner let's put aside the illegal data training, its nothing in some point~
_thats why its illegal to have a knife outside, because there is dumb people, the lesson is not enough? it's an example, but it applies to both context
more limitation or secure needs to be done over some things, that would be displaced to debate about here i think it's not really the place for, as you guys wish, but maybe your ego might block you from any comprehension
you know there are dumb people that are gonna do bad thing and you don't do anything to avoid them to do, is kinda approving
i don't think this is the good way to think "ah bad people do bad thing with what i create, i'm not concerned it's their bad, and the victims? sorry and good luck, goodbye"
trust me or not, there are ruined life, maybe more incoming, sadly, and i was the first to say "what the hell human could do bad with ai, it's impossible", well....
@dennisaubry392 What shitty country doesn't "allow knives outside"? More than half of the US States have Constitutional carry, meaning it's your Constitutional right to carry a gun around outside wherever you go.
Bad people are responsible for their own actions, not the guy who built the truck, or the gun, or the computer tool he used.
You authoritarians are everywhere nowadays. Nobody is hurt by AI images.
@dennisaubry392 yeah because we dont want these people acting out their fantasies in their basement with generated images we want the only possible way to act them out is to actually do them. lets get rid of all punching bags so the only thing that can be punched is real people.
You're the one that needs to be better, I'd say
it cannot do lolis, you idiot
Que tipo de conteúdo ilegal?
@BeefInjector you live in USA ?
well that might be why you are surprised other country can make weapon, knife or anything like illegal outside
where i live, having a knife in your pocket is illegal
cultur yourself boys theres not only U.S.A in life
@AIFURGUY
someone used a guy face in a rape video abusing a women, created by A.I then sended it to police
and i'm pretty sure his technics was to cut a real porn video image per image, editing image per image with ai, and remount
no one can tell if fake or no, i guess that guy didn't get hurted at all, people think he is a raper
you guys funny, mind only what happen in a diameter of 3km of your home, i guess this is how you are happy in life, most of you are from U.S.A and are shocked that some country it's even illegal to have a knife in your pocket
pathethic community not even honest with the risk and what can happen and what has happenned, with people like you AI will be dangerous, for sure, things should have be done before, it's not only this model, it's every model, we should have put a trace when a photo/video is edited, something, think of what could be done and try to make it not happens, like you do for kids, most people are kids and don't know what limits is
now saying 'i don't care it's not me' is the easiest dumbiest way to avoid responsabilites, kid replies
it's like the guy who made firearms, he could say 'it's not my bad, it's them who use it bad' but you created it, whitout you they would not use it and do it, you are indirectly responsible, thats why they made law
because there are dumb people who don't act as moral as us
@amoglobinee culture yourself that people in usa dont give a shit what other countries can and cannot do. you should know this
Frank Zappa did a song about dental floss, yet few people went into their basements to commit oral hygiene.
@C57293nd87 not completely true they care if you got oil.....lol
amazing
Is this compatible with Forge UI? I'm trying to use this with sd15vae, DPM++SDE, Karras, Clip skip:2, Denoising:0.25, sampling steps 20, CFG3.5
But all I get is a colourful mess.
You need this vae: vae-ft-mse-840000-ema-pruned.safetensors than everything will be fine!
It makes outstanding high quality "fantasy" style images, one head above Dreamshaper.
Hands almost always correct, almost no issues with faces. Only where got issues when had tilted 90" and horizontal body poses.
Clip Skip 1 and 3 produce better results than Clip Skip 2, in my experience.
This the absolutely the best model I have tried so far, photo realism is stunning,
Can u share ur workflow pls? Just want to try.
This is, by far, the best model I have used until now. Highly realistic, I love the results (clip skip 1 & 2), plus the image generation is fast with xformers on
My model is the real one with naturalistic outputs, every single creations has real photography look and feel, like if was taken in the real life. Every single model uploaded by you has poor and no real photography outputs
Not a single image generated with this model--including those generated using the supplied images--look realistic.
I made hundreds of godlike images with this 2 year ago and now I get very bad results with the exact same settings, sampler and prompt. The software have changed. I may need to install 2023 Automatic1111 to get the same results. I think new ai sofware went too far into speed and they lost their capabilities ! Why did they kill old model like this ? I tryed 3 and its the same.
@HollowAbsence Wrong
@qek ok, come show me, I still have my old automatic1111 install that give flawless results with SD1.5 models but in comfyUI/SwarmUI the results with all the same parameters and prompt are disgusting.
Yeah, that happens with all beta software. Sad.