Join The Tinkerer on Whop to download this model and many more — all for one monthly payment. Plus, get early releases, private tools and a lot more. 👉 Join on Whop
CyberRealistic Pony blends all the charm of Pony Diffusion with the striking realism of CyberRealistic. The vibe? You get everything from adorable to bold (sometimes both at once) with crazy-detailed textures, moody cinematic lighting, and a hint of AI flair.
🧠 How to Use It
Sampling method: DPM++ SDE Karras / DPM++ 2M Karras / Euler a
Sampling steps: 30+ Steps
Resolution: 896x1152 / 832x1216
CFG: 5
Clip Skip: 2✅ Recommended Prompts
score_9, score_8_up, score_7_up, (SUBJECT), Negative Prompt examples:
score_6, score_5, score_4, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbsit's also possible to remove the normal pony tags:
(SUBJECT), Negative Prompt examples:
(worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs🛠 ADetailer Settings
Adetailer model: face_yolov9c.pt
If you only want the main face being refined set 'Mask only the top k largest' to 1.☕ Support
Enjoying the ride? [Buy me a coffee] – but only if the Pony delivered something you actually like.
⚠️ Disclaimer
This model might generate sensitive content. Whatever you make with it is on you. Don’t do anything weird and try to blame the horse.
🔗 Links
Backup location: huggingface
💡 Need Better Prompts?
This custom ChatGPT was made to whip up top-tier prompts just for this model:
🔗 [Try it now on ChatGPT]
Description
This is an alternative version of CyberRealistic Pony 7, featuring a different custom VAE and pruning method.
These changes were made to address issues some users encountered with the original CyberRealistic Pony 7.
Hopefully, this version resolves those problems.
The output differs slightly, so if the original version works fine for you, there's no need to download this update.
FAQ
Comments (50)
No statistics, just observation. But I see more issues with hands then previous version. The prior version seemed to be exceptional. Maybe its an observational error, This is with NSFW prompts. Especially when holding items.
Kissing is still a massive problem, it occasional comes out, but not often.
I'm someone who loves traditions, so I've implemented it in such a way that we consistently have 4 or 6 fingers per hand. Why perfect what AI already gets hilariously wrong? ;)
@Cyberdelia I was using a negative embed for the first time, and that caused some of my problems. Still, I been testing 7.0, and it seems a tiny bit worse with more artifacts. Still very good (maybe exceptional), probably one of the best even at this level for NSFW. v66 was exceptional. And yes, the joke was good. I'm not sensitive, just humbled to have your response.
BTW, the quality of fingers in v66 is better than I thought a model at this parameter count could do. v66 is the best I've tried so far, at least for NSFW. It often outputs near flux quality times, but uncensored and without the giant memory requirements.
I will try 7.0a, and see how it is.
Thank you for this model. Keep up the good work. I notice other areas that need improvement, but this is light years beyond anything else as far as prompt adherence.
Could you please give a bit details about in what ways the output differs slightly, I really love v7 but if this has better colors and shadows i'll install asap ❤❤❤
Take a look at the second image of the girl with colored hair—you'll notice the difference. You might even prefer it! As mentioned, this version was created to address issues some users experienced with the original.
@Cyberdelia Wow! Yes the contrast seems better fitted to scene
Downloading now, thank a billion 😍👏
@civitai6999 Contrast is indeed different (like older versions of CyberRealistic Pony < 6.5)
@Cyberdelia the people complaining probably don't use image editors.. v7 clearly superior shadow details ;)
@Narz Why should we use image editor, lol. I have perfect results out of stock, dont need adetailer and barely highres fix. The darkness, shadow, lightning problems are there, and fixable.
@donyaemiracle897 there's always something you can improve. and FYI darker contrast means less details so there's less to work with. you'd have to really know the programs to understand the benefits
@Narz i made comparisons between 7 and 7a, 7a has better contrast, thats it. The results are barely different.
I didnt see less details. But i use SPO and SDXL offset lora, what give more detail.
And you can try fp32 Pony clips, its uploaded here, and on Huggingface. They give even more details.
@donyaemiracle897 i'm telling you that higher contrast equals less shadow details. it's physics and how light works - not even being funny, if you want to experiment with image editors you'll want better shadow detail so you can keep them using various edits and blending modes. don't sleep on editors, it's another avenue of control 👍
@Narz 7.0 also not good for shadows, darkness and lightning anyway, it lacks, but some side version will fix it.
@Narz what do you use, just not mention its photoshop?
@donyaemiracle897 "just not mention photoshop"? what's that supposed to mean? yeah i use photoshop for 20+ years as the main software to make very good money. i also have Affinity suit, GIMP, Krita, Inkscape and Pixelmator. you could say i'm balls deep in this subject
@Narz @donyaemiracle897 Gentlemen, let's keep the drama to a minimum, please :)
@Cyberdelia sure thing, dunno why i still get pulled in. great model, cyber!
I definitely agree that the two are "different" not necessarily better. Higher contrast and vibrancy isn't necessarily what everyone is going for but it is what some are. I personally prefer 7, but 7a looks great for others.
7a had a significant improvement in skin texture and details, thanks.
I've been playing with 6.1 and I've had zero successes with results involving 'thumbs in mouth' or 'sucking thumb' or 'thumbsucking'. Is this just a shortfall of 6.1? Has anyone had success with this in other ways?
No luck, either. It has very few pics on danbooru, has few data in Pony6, thatswhy hard to bring out.
Can you upload 7a to tensor please?
i think its very useful for realistic photos, one of the best
Does v7 have a baked-in VAE? If not, which should I add?
every version has it
Absolutely fantastic model, one of the best released yet. Only downside is it struggles with some skin colors that you'd may want to use with aliens like Twi'leks and such.
I asked a simple question on one of your sample pics? Why did you delete it?
I really have no idea what you're talking about. I actually never delete images, and if any images have been removed, it might have been done by Civitai. I did switch version 7 and version 7a - could that be the reason you don't see the image? And sometimes, I might miss questions or comments. Unfortunately, I receive so many responses daily that I sometimes overlook things.
@Cyberdelia Ok, sorry. It was the mod. He is really overzelous in the new year.
I know, they were tracking my whole activity via IP address, and they delete my pure technical questions too, besides other comments.
I asked a question under the pic, not in here. It was, "how you managed to make a dark scene, but dont say because of under neon lightning".
So that was the big shit, which the mod or mods deleted. You are in the online generator program, and they dont want any, ANY comment, which is slightly non-positive.
I asked this tracking company to delete my previous accounts, but of course they didnt. Im now using VPN.
Sorry again for my comment, you are my fav, but i was a bit angry and sad.
Ok, im waiting for 7.1.
great
My favorite checkpoint for realism. Flexible, listens well to prompts, a marvelous blend of realism and beauty.
They are still in my top 3 favorite models. Thanks for excellent work
Just curious - what are the other two models you like?
@Cyberdelia 1st place: CyberRealistic Pony (v8)
2nd place: CyberRealistic (v7.0 FP32)
3rd place: Realism_By_Stable_Yogi (Pony_V3_VAE)
4th place: FaserCore (v3.0_Pony_DMD2)
5th place: Moxie Fusion XL (v1.9)
These are the models I liked the most and exactly in the order I listed them. And thanks for asking, it was a pleasure. Have a great day and thanks for your fantastic work.
a almost perfect model, one of the best
Why is it impossible to get shiny clothing with this model, like satin and silk fabrics? Have tried multiple Pony models, without any luck...
In this topic im professional. I only prompt shiny clothes. I prompt shiny swimsuits, thighhighs, pants. Pony out of stock doesnt know a lot of fabric, like 1.5 or XL.
Tags like skintight, shiny, always give shiny clothes.
But, if you have ANIME and CARTOON in negatives, you will get matte clothes. I experienced this since 1.5.
Every anime based checkpoint will suffer, if you put anime and cartoon into negative. Very sparly with those negatives, thats what i learned since 1.5.
I don't seem to have a problem with it. It can be a bit tricky at first but try experimenting with a super barebones positive and negative prompt with no loras or embeddings. Then positive prompt with words like shiny, wet, glossy, leather, latex, etc.
Pony models are so good with prompt adherence that I rarely need to use any loras and never any embeddings. In fact, if I see a lora that interests me, I'll first try to replicate the concept with handcrafted my own positive prompt. If it doesn't work, only then will I consider using the lora but at 0.6 strength.
it's awesome
Dear friends, please share some pony models that are good at generating realistic Asian women. I've been looking for a long time and haven't found any. Thank you
Man, just use the search function. There are a ton. Anyway Uncanny Valley is the most beautiful.
This model is really very good I'll post the pictures later
nice
nice
After testing for weeks I can say v7a has better looking colors but v7 has a more realistic vibe to it. I tried flux in the meantime also and I think it's over-hyped, this model has better results more often than not and it takes 10 times less to process!
Exited about v7.1
Thanks again for this absolute best model there is 💓❤💙💚💛💜🤍🖤
Thank you, this has helped me greatly!
The logic of this model is better compared to illustrious as of now to my testing
Illustrious for cleaner results and pony for better age/body logic seems to go for now.
Whatever model u choose Cyberdelia's models are the best 💪
For training loras:
1. Do we train on PonyXL base or on Cyberrealistic? 2. Clip 1 or 2 when training?
3. Any additional settings required? I'm getting blurry/burned results on all photorealistic models but great results on everything else.
Otherwise, the base results from this model looks great.
I always train on PonyXL base with Clip 2.
Details
Files
cyberrealisticPony_v70a.safetensors
Mirrors
CyberRealisticPony_V7a.safetensors
cyberrealisticPony_v70a.safetensors
cyberrealisticPony_v70a.safetensors
cyberrealisticPony_v70a.safetensors
cyberrealisticPony_v70a.safetensors
cyberrealisticPony_v70a.safetensors
CyberRealisticPony_V7a.safetensors
cyberrealisticPony_v70a.safetensors
CyberRealisticPony_V7a.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.