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
Let’s go for visuals that don’t look like pony selfies, kick those bugs to the curb, and add a dash of more realism - what more could we ask for? :)
This version includes new realistic training that I had partially used for CyberIllustrious.
FAQ
Comments (146)
Do something about the horse's teeth. Describe what's new, the 7th version is generally not bad, how is this better than it?
Looking at himself in the mirror with his buck teeth... what's wrong with buck teeth? :)
This version includes new realistic training that I had partially used for Cyberillustrious.
@Cyberdelia I believe someday you will get something better than version 5)
@labrombula378 the golden version :)
@labrombula378 I could try to add these new training on Version 5.... mmm.... maybe tomorrow (upload it on huggingface)
@Cyberdelia The best is the enemy of the good)
@Cyberdelia I would love to see a new improved version 5!
Testing V8 and it's very nice, in my opinion, more realistic than V7. Especially the hands, much more consistent. A step forward.
Great Job! I love all versions 😍
And I love deeznutz!
WOW! Beautiful checkpoint!!!
After playing with Illustrious models for the past weeks, it feels like coming home to this model, and this version is just amazing. Going to use this model a lot until Pony v7 is done cooking.
V8 Love this version, getting great results
I'm a simple man. I see a new CyberRealistic_Pony model, I download. The results look very promising so far. The babe agree :D
very nice
Why do I keep getting "safe_pos" in my prompts every time I generate today, and in the negative prompt "civit_nsfw, safe_neg"?
I don't understand. These things spontaneously are added and that's in the onsite generator? Guess it's a new policy?
very nice
Will v8 be made available for on-site generation?
I have asked Civitai to swap it from V7 to V8. Unfortunately that's the only thing I can do
@Cyberdelia ok, thank you!
You really turned up the realism in this edition! Thank you for an excellent version 8 and introducing me to "open wool knitted full bodysuit" 🙏
Loving it so far
Can somebody tell, what canged in V8?
I use many detailers, and what i saw only, V8 had bad eyes and less detail.
Great Checkpoint! works like a charme!
I really hope we eventually get V8 available for generation on civit
It's finally now online
amazing model with amazing results!
Amazing model as always. The realism is off the charts
Will you upload it on Hugginface? I am not able to download from here :/
Please add the generation option for v8 on the civitai. Thank you <3
Go crazy - it's swapped now from V7 to V8 :)
@Cyberdelia I have already managed to train LoRa for the new version 8, so I would like to refer to the new version of the model in my works. 😋
@blacksnowskill nice!
Im getting NaN errors sadly again :(
back to 70a for now I guess
So weird - I will try to make a 8a version and drop it on huggingface
@Cyberdelia apprected GL please let us know
Same here. The VAE isn't working for me either
any update?
I've been testing this; switching between v6.5, v7 & v8 back & forth. All the results are surprisingly similar using the same data/seed (this isn't usually the case with other versions I test across for different checkpoints). v8 is coming out slightly ahead as superior though (vanilla) - slightly less busted features, eyes, anatomical compositional goofs (quite subtle to generally detect sometimes, but you just get this uncanny valley feeling with something just being a bit off more in older models, even if it's not outright big goofs with fingers, etc).
v6.5 was my fav of the bunch prior to this one. I feel confident this is the go to model now, enough so to delete my older versions. v6.5 usually made more anatomy errors, but there was just something special about it's aesthetic/style that I didn't want to let go of until now.
https://civitai.com/images/54932204 v6.5 (my fav, but when prompt complexity increases, the checkpoint starts to buckle under the strain more compared to v8 (bad hands/anatomy). Differences tended to dimish more the more complex I made the prompts/more akin to real prompts I'd actually use with all the add-ons).
https://civitai.com/images/54932258 v7 (notice the busted eyes)
https://civitai.com/images/54932358 v8 (the most realistic probably, but I find it just lacks style compared to v6.5 a bit). v8 is way more consistent/reliable as prompt/input complexity increases though, so v6.5 just becomes frustrating to use/fix under real working conditions by comparison.
This is a tiny sample, but I did this many times over. This is a generally good representation of base outputs I found. I typically avoid LoRas for testing, because a lot of times it's not reflective of model strength, but rather idiosyncracies between them just clashing together badly).
Yeah, interesting and thanx for comparison. I use many detailers, thatswhy i had almost zero change in V8. Actually it has worse eyes and less detail. Realism doenst progress anymore, thats the max.
Great effort!
Развивай этот проект,
10\10
V8 - Great model. Tested against previous versions. Only failed one hand test (but don't they all eventually :) ) the realistic results especially on some of my non person tests like buildings, furniture etc. were definitely the best (looked at my 6, 6.2, 6.4, 7 as compares)
Best tip for more realism to use dmd2 lora. The result is 10-15 percent less detail, but way more realistic and at least double speed. So it worth it.
can you share full name? I only found this ''Experimental Pony Real Dream Realistic Lora'' and it doesnt improve speed even 1 second
https://huggingface.co/tianweiy/DMD2/blob/main/dmd2_sdxl_4step_lora.safetensors
You can use within range CFG1-2. Best setting seems 1.5-1.6. Use with LCM sampler and Karras or Exponential scheduler. Steps for best quality and detail within range 12-16.
You can use with Euler a too, sharper and more detail, but can be artifacty.
About detail loss, it loses detail on eyes, and hair(if you prompt fancy hairstyles with multiple colors and styles). I advice to use detailers (the best ones are SDXL, but they work with Pony).
Realism much better, but you will lose those fancy waifu eyes. If you dont care about this, go for it.
@luisenriquejoby839 Thanks a lot for your time, tried this one however the speed is still exactly the same, could you please tell which UI are you running?
edit: Nevermind, tried with the LCM karrar and it indeed speeds it up 2x. But the loss of detail is a lot more than I hope for. Still thanks for teaching this method! Have a nice day!
hey may i know why im getting nude images even if i didnt put that in the prompt??
I know, isn't it great?
@Cyberdelia bet this is the same guy on reddit
@civitai6999 lol
Not exactly great when you show your grandma how you're making images with AI, right?
@sent_data_proc I’ve got a feeling you’ve landed yourself in this pickle, and I bet you’re just dying to spill all the scandalous details. I’m grabbing the popcorn already!
@sent_data_proc Even better!
Very quickly became one of my fav models! Well done mate :)
all of ur version ,cant even do small males,no matter what,all girls
Yeah, my model refuses to make small men. It keeps insisting ‘Tom Holland is just a state of mind’ and auto-corrects every Danny DeVito into a smug anime gremlin. At this point, I’m afraid to type in ‘Joe Pesci.'
Hm, i had no time to give a chance for 6.6. It has the best eyes and beauty, but too contrasty.
Now with v8 i wasnt impressed with eyes and look. I tried many prompts, and with "aesthetic" it seems okay. Realism way better.
V7 release before christmas was so joyful, i didnt even compared to previous versions:)
My take is 6.6 for beauty and v8 for realism now.
awsome
so cool!
Version 8 is perfect, as are all previous versions. Thank you so much for the great work and I pray that you will continue with PONY and SDXL models.
https://huggingface.co/Mescalamba/CybeRealistic_PONY_v8_GGUF
If someone wants to use GGUF version, its there along with all stuff needed. Pardon the typo in repo name. Hope its okay with author.
In case you wonder "why", answer is there.
Very good, nice.
I would mention the importance of fp32 clips, much better quality, than default fp16.
If you use Forge, than simple write in webui-user.bat --clip-in-fp32
If you use other gui, than maybe you need to download the fp32 clips, which are here on Civitai.
@khalanibroden426 It wont really work, cause PONY needs PONY clips, only option is to merge them with better ones, but it doesnt really work great.
-fp32-text-enc enables same for ComfyUI
@Mescalamba I dont know, but Forge seems to have built-in clips. I only used the command and the results are way better.
This is for the basic model, not your fp8.
@khalanibroden426 that GGUF is fp16, not fp8, its full model, just wrapped in GGUF
Not sure how Forge works. :D
@Mescalamba I think Forge can force basic SDXL fp32 clips.
@asadbekdevonta998 Yea and it wont work, cause PONY requires its own CLIPs. If I dont mind they need to be plugged in reverse.
Version 8 is absolutely amazing. I did not imagine that something can be so beautiful!
I’ve been using this model recently and it’s great! However suddenly right now I can’t use it as it shows a message that says “Some of your resources are not available for generation: V6 (start with this one)” could I get some help please?
same here
I have the same issue. Worked fine until a few minutes ago.
it came baaaack! woohoo
@DIXTOXTUDXO Same here!
Good stuff. Adheres to prompts well enough.
Someone has an ultrarealistic workflow for Forge?
I know Refiner(slow and shit), highres(slow, not good enough), Adetailer(slow, and only for eyes).
I could use Photoshop workflow too.
For this model its hires fix (or just refiner, which is basically the same).
If you want easy and somewhat fast realistic stuff, get FOOOCUS (dunno how many Os should be there) and plug in some realistic SDXL model. Or some fork of that software. Its old, but probably easiest to make realistic pics with.
@Mescalamba Why its more realistic? What do you mean about xl plugins?
I use CyberPony, because i have a ton lora, and its the best balance between compability and realistic. I would use porn XL models, like BigLove or Lustifiy, but none of them has compability.
Very nice..
swag
cool
So cute
Tony Stark here to give this model a 9/10, very realistic
good
Peak
One of the best, hands down. However it has issues with img2img, perhaps an inpainting model would be useful.
One of the best Realism models for Pony, very solid results.
Good one
really good!
This is f'ing amazing - best realism mod for Pony.
hi
How does it compare to Realism by Stable_Yogi?
嗨
waaaaay too good
Just downloaded the model, didn't even do any work yet, but even after just a few first technical tests I'm very impressed with this model. It is the only Pony-based model I tried so far, which was able to upscale the image x2 without going bonkers. All other Pony models I tried can only do an x1,5 upscale. Secondly, this model handles inpainting like a native SDXL - most Pony models are only able to slightly change details without losing it, this one actually inserts objects into the scene with seemingly decent fidelity. I expect the model to be really great with producing clean images with minimal AI hallucinations. I will be testing for prompt adherence and limits of capability, but already I'm deeply impressed by how good this checkpoint is. Respect!
shbgtdnhtd
They seem to give buzz for comments again. People dont even try to write something, just spam the models. Daily beggars to generate some pics.
Yes, terrible :(
@Cyberdelia Be happy!:) They spam only the top tier models and new pics to gather fast daily buzz.
@sitiveniexton261 oh yeah 👍 spam on dudes!!
Love it
Very good
Im gonna be honest. One of the if not the best model for body and facial details. Especially on the eyes. Tried so many other models just to see how much detail they can produce, and this stands alone in the whole community. Gets even more detailed and perfect with upscaler and inpaint when used for this model. Hats off to you sir. You truly achieved some top tier work here!
Hm, what v8 lacks is exatly good eyes. I use detailers, so it doesnt matter, but the pure model lost the beauty for realism.
But i try a lot models, than back to Cyber:) Best balance between adherence and realism.
it is way good indeed.
This is hands down the best model I've ever used to date. Massive props for the insane work you've put into this! 🙌
Have you tried this model?
https://civitai.com/models/1291404/dhxlv10
Basically, like Pony and Illustrious, but there are inbuilt realism. I think, this would be much better for further realistic finetunes.
The realistic style(style_3) is almost on par with realistic Pony and Illus models.
Masterpiece!
Allright. Every model got update at once. Where is Pony?
Very good !
Nothing realistic about it. Screams plastic, anime, render.
Thank you! I was going for 'Barbie meets Final Fantasy at a wax museum' vibes. Glad it came through! 😂
Man, beside some XL hybrid, there is no realistic Pony or Illustrious. But XL hybrids sacrifice the natural adherence and have incompability with characters.
You use either XL hardcore models(very few good and only 1girl generator) or accept that this is almost the maximum realism without sacrificing, what is Pony.
And put those tags into negative:)
@Cyberdelia I would subscribe to your comments alone if I could.
It happens sometimes, this is because prompting. The prompts you may be trying to use together are not vibing with each other. And so fake looking elements seep in and so everything looks fake. If you use prompts that vibe together or are "down to earth" you'll get good restults. Additional I do notices pony can give you shiny skin, put shiny skin in negative if you want.
Love this model but feels like something changed with it recently. It constantly tries to put penises, sex, and penetration into images even with no references to them whatsoever. I frequently have to massively overweight negatives to keep them out - even heavily emphasizing "solo" in the prompt is not enough to correct this. Never used to have this problem. Very annoying.
rating_safe
But the problem is on your side. Im heavily into hardcore, and must emphasize the sex act to not get solo girl. Ok its a pose, which fails on almost every realistic model.
Please keep updating this model, I really like it :)
Where can I find v.4 and v.4.2?
https://civitai.com/models/15003?modelVersionId=372799
Was there Pony 4.2? i dont know.
@cheskykaras616 What you linked is not the Pony version, that's SD1.5.
v4 of CyberRealistic Pony can be found on Huggingface.
@mxnai found it! Thanks for redirecting me.
@JamesMccay1995Ai Ah, a fellow hoarder? Or just someone who can't resist the urge to have it al :)
@Cyberdelia No sorry, not a hoarder. I had created some old images with v4 and v4.2 that turned out really good. The latest version were not creating the same composition as v4 and v4.2 did. But anyways, I ended up finding it on an archived github page. All good :)
You should release the new version soon, just for the sake of top 6.
Thats two row:)
Haha, luckily I don't get influenced so easily by these kinds of things. The new Pony version isn't far from release. Hopefully mid-March, but possibly sooner.
Update: maybe way sooner. Training was done correctly (just ready) so put things together and release it
@Cyberdelia Thanx man. While Illustrious is good, Pony is still the most popular.
Guys, is it possible to add this model in Fooocus?
Of course
use it with eular ancestral, any dpmpp samplers tend to make images look really rough i think, well they do in foocus anyway
There are some youtube videos on how to use pony model with Fooocus
Tried to add but it crush
@joooove596 it works normaly in my system try use pony persistem 1 time and make sure you have good ram 32 gb could crash in 16
@rishabhsinghcoc01854 ram 32, vram 8.
I put cyberrealisticPony_v8.safetensors in \Fooocus\models\checkpoints
styles > all off
Models > base model - cyberrealistic, refiner - off, lora - off
And nothing. Just crush
@joooove596 what are your CFG and sharpness settings? and how many steps are you using
@mariedoll123 cfg 5 steps 30 as recommended
@joooove596 its possible something may have screwed up during the install, have you tried deleting foocus and installing again ?
@joooove596 try reinstalling it maybe some files could have mismatched and try downloading it from different source
@rishabhsinghcoc01854 no never download from a different source , only download from the official site .. https://github.com/lllyasviel/Fooocus
@mariedoll123 even i have downloaded from here, you can watch some youtube tutorials and follow each steps maybe that could fix it//
Details
Files
cyberrealisticPony_v8.safetensors
Mirrors
cyberrealisticPony_v8.safetensors
CyberRealisticPony_V8.0_FP16.safetensors
cyberrealisticPony_v8.safetensors
cyberrealisticPony_v8.safetensors
cyberrealisticPony_v8.safetensors
cyberrealisticPony_v8.safetensors
cyberrealisticPony_v8.safetensors
cyberrealisticPony_v8.safetensors
cyberrealisticPony_v8.safetensors
CyberRealisticPony_V8.0_FP16.safetensors
cyberrealisticPony_v8.safetensors
CyberRealisticPony_V8.0_FP16.safetensors
cyberrealisticPony_v8.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.