If you want to use more my checkpoint online generation, please visit here.
https://tensor.art/u/762555264535746522
You can run WAI-ANI-NSFW-PONYXL and use its API on SinkIn:
V14 released.
Increased composition diversity and improved model accuracy now allow images to be generated with fewer steps.
Recommended Setting
Steps:20-30
CFG scale: 5-7
Sampler: Euler a/DPM++ 2M Karras
ADetailer face_yolov8n/s.pt use can fix eyes
For the example images
I used 1024x1360,
generated directly without AD or hires fix.
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,Negative Prompt
worst quality,bad quality,jpeg artifacts, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark,V13 released.
Increased body stability and accuracy.
overall balance adjustment.
Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a/DPM++ 2M Karras
ADetailer face_yolov8n/s.pt use can fix eyes
For the example images
I used 1024x1360,
generated directly without AD or hires fix.
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,Negative Prompt
worst quality,bad quality,jpeg artifacts, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark,V12 released.
Style adjustment, background adjustment, and increased stability.
Added more compositions.
Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a/DPM++ 2M Karras
ADetailer face_yolov8n/s.pt use can fix eyes
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,Negative Prompt
score_6, score_5, score_4, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark,v11
Weight adjustment, balancing the art style
All example images(Except the cover) are generated at 1024x1360, without using AD and hires fix.
Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a/DPM++ 2M Karras
ADetailer face_yolov8n/s.pt use can fix eyes
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,Negative Prompt
worst quality,bad quality,jpeg artifacts, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark, V10
Better Background
Clothes details up
Better mature female
Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a/DPM++ 2M Karras
ADetailer face_yolov8n/s.pt use can fix eyes
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,Negative Prompt
worst quality,bad quality,jpeg artifacts, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark, ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
V9 Hyper12step
Please refer to the instructions on the right side of V9 Hyper12step for usage.
V9 released.
v8->v9
adjusted the facial data to make it more versatile, allowing for easier generation of faces across various ages.
added more NSFW materials.
attempted to reduce the generation of some unnecessary details.
Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a/DPM++ 2M Karras
ADetailer face_yolov8n/s.pt use can fix eyes
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,Negative Prompt
worst quality,bad quality,jpeg artifacts, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark, __________________________________________________________________________________________________
hyper12step released.
about size
Various combinations from 768x768 to 1360x1360 can produce good images.
Recommended Setting
Steps: 12
CFG scale: 3.5
Sampler: Euler a
Not very necessary to use ADetailer correction
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,Negative Prompt
score_6, score_5, score_4, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark,PS.The example images use a size of 896x1192. no AD no Hires. fix
__________________________________________________________________________________________________
Version 8 released.
v7->v8
Added some training related to camera angles and backgrounds.
Added more NSFW materials for additional training.
Added some training for special effects.
Optimization of overall composition logic.
Fixed some minor issues.
better eyes and hands.
Recommended Setting
Same as v7
__________________________________________________________________________________________________
Version 7
v6->v7
Fixed the issue where images appeared entirely covered with snowflakes.
Revised the composition logic to be more suitable for general scenarios.
Added new assets for backgrounds, mature women, and clothing.
Ensured consistency with the previous model and further fixed the issue of blurry eyes in medium and long shots.
Fixed some minor issues.
V7 Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a/DPM++ 2M Karras
ADetailer face_yolov8n/s.pt use can fix eyes
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,Negative Prompt
score_6, score_5, score_4, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark,________________________________________________________________________________________________
Regarding the 8step model, it is a lightweight model that can generate images in as few as 8 steps. However, the quality and text comprehension will be relatively lower. If speed is not a major concern for you, it is recommended to continue using version 6.0.
8 STEP Setting
Steps: 8-12
CFG scale: 3
Sampler: Euler a
ADetailer face_yolov8n/s.pt use can fix eyes
Positive Prompt and Negative Prompt same 6.0
__________________________________________________________________________________________________
6.0
v5->v6 Change point
Better Background
better eyes
fix vae
Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a/DPM++ 2M Karras(better eyes)
ADetailer face_yolov8n/s.pt use can fix eyes
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,BREAKNegative Prompt
score_6, score_5, score_4, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark,You can view the Supported art styles from https://civarchive.com/articles/5715
Thank you, DeepDark_Fantasy514, for providing the test.
v4->v5 Change point
Better Background
Clothes details up
Add some mature female
Little better hands and eyes
Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a
ADetailer face_yolov8n/s.pt use can fix eyes
Positive Prompt
score_9, score_8_up, score_7_up,Negative Prompt (THX for snipsnapsnipsnap recommend )
Base:score_6, score_5, score_4, source_cartoon
Optional: source_furry, source_pony can bleaches skin
else: 3d, (censor),monochrome,blurry, lowres,watermark,
I often use it now
[score_6, score_5, score_4, source_cartoon,
3d, (censor),monochrome,blurry, lowres,watermark, ]
Description
FAQ
Comments (59)
This is my favorite model.
it is too but then I tryhard learned the ways of vpred and the broken details lora.
it is too but then I tryhard learned the ways of vpred and the broken details lora.
Hello,
here's my quick impression of the new version.
v12 worked really smoothly with the internal generator. The consistency was really nice and the prompt was mostly implemented properly.
Unfortunately, everything is really unreliable with the latest version (v13) and the male faces in particular are really not good. The eyes are sometimes significantly different sizes and shockingly bad compared to the previous version. The bodies are completely inconsistent and the optics are very blurry.
Got it, I'll try to fix it in the next version.
Imagine ever having a male face in frame ever
This model you just whip out Euler a karras 30 steps and go 1024x1360 with your “oh my god so frustrating 9_up etc” first line of quality text and it spits out an image 9 times out of 10 better than the front page of ai booru. I really do think IL people are coping although I do see the fun in noob stuff with predv. But it’s like learning an entire new language. If you don’t use artist names and format the text perfectly it looks like shit. The only time I get anything not looking like complete shit is when I rip somebodies meta data just to see if anything fishy is going on. Did it to that sword girl on obsession noob xl (superior to noob imo) and got the exact picture just less detailed and I yes I did follow all the special meme rules you have to learn. All I’m saying is this model never lets me down. Oh no I might have to inpaint a hand for an hour. Happens on every model. The problem is this looks legit professional good without knowing every single artist alive and every single little nob to turn. Ironically it goes in the other way too. People who are just okay with having no background shit art love IL because less words. “Goth gf who love me forever 2024 best quality.” And you get an average looking anime girl. Thats probably 99% of the hype because aesthetically unless there’s forbidden knowledge out there somewhere most IL looks like shit. Also Euler a karras master race.
i have been doing this a while cant believe i never tried this "Euler a karras 30 steps" thanks
@DudeNessNess I use vpred now so no more karass
@DudeNessNess literally just copy people with the best pictures who keep their meta data on it. That’s how I got remotely not shit at this hobby. And probably everything in life. Emulate then you can mess around.
Can't live without this model.
How can i use this model to make hentai of mavuika from genshin impact for example? Or other famous characters from games/shows, just giving mavuika because she's very recent
it's a good thing google exists. look up "stable diffusion"
look up comfyUI and watch a tutorial
can use the webui on this site ("create" in the top right), or comfyui/a1111 to generate offline/unlimited (will need at least git & latest python installed, pip & pytorch if you want to do more advanced gens)
just using the site is a good start, not sure if this specific model is available, but the model it's based on is
Two different versions of your model, but version 13.0 did not become better, but became much worse. I do not know why on version 13.0 and 12.0 some lore models do not want to generate AI art well. So at the moment I recommend using version 11.0, as it is more stable and compatible with most lore models.
V11 was the best version so far, yet, owner doesn't care about allowing it to be used on site
This model is very excellent!
I feel that the combination of "grabbing another's breast" and "pov hands" does not work well. If you could add to this model that would be great.
I like the artistic touch it gives to my pictures. They get volume and shades
Great, respect
it does has more body stability and able to try new prompt for it, can't wait what v14 looks like, thank you for your hard work
Great model in almost every way, but hands are still absolutely terrible. I can barely use a third of generated images of humans, because the rest has totally messy hands.
I am kind of confused by how this model works. Currently I am using v11 (I was using v8 some time ago, right now the latest is v13).
My generations always lean more realistic than anime. Like a painting or an American comic/cartoon instead of anime.
From my testing using BREAK makes things appear more anime, the more BREAKs the better. But there are also images I find that use no BREAKs at all, yet the images appear to be in an anime style. I've taken an image that used BREAKs, removed the BREAKs, the result appeared very similar (style remained anime and did not become more realistic/cartoon). Other times I took a prompt that gave me realistic images, added BREAKs, got more anime-style results.
So really confused. Guessing that the generation of part of the prompt affects the style of the character if a BREAK isn't applied, but I am no AI expert whatsoever.
Edit: I am also guessing the size and order of prompts affects the style of the art.
Otherwise a good model so far (on v11).
Update: I downloaded v13 and started trying it out. Its style is definitely a little different than v11. While I think v11 leans more cartoonish-realistic, but still with heavy anime style, I'd say v13 is more 2D anime and much less cartoon/realistic. v11 is a little darker with more "painted" appearance. v13 uses brighter colors and a little less color changing on a surface.
I'd say v13 has a more anime feel. But v13 isn't as compatible with some loras I use. In one test, v11 created better results than v13. If I could attach images I would.
I'm curious about how the results differ for previous versions but am not bothered to test and compare every single version.
So for now I'll keep between v13 and v11.
Just put realistic in the negative prompt and in your notepad that has the long line of opening prompt text (9_up masterpiece etc) just add “source_anime” in there. This model really doesn’t have that issue from what I remember only coco models usually do. Just follow prompts from my images if you want I keep meta data on to help others.
Also are you using an external generator? Don’t think I’ve used BREAK in my entire life and that only works on webui with certain extensions and even then it’s still default /n and not break. Honestly if you’re just learning I would recommend just not using pony and learn noob/vpred.
Could someone recommend me some loras to fix anatomy? I keep getting a bunch of generations with multiple/detached limbs.
You’re probably doing hires fix with like .7 denoise or upscaling right? If you just create an image 1024x1360 with the suggested quality text (9_up masterpiece etc) and negative prompts you’ll get great pictures with this model. Upscale only 1.5 at a time max .4 denoise and use controlnet if you don’t want the picture changing much. If you’re talking about hands/fingers/toes you’re just gonna have to get lucky. Models this refined rarely spit out multiple arms unless you’re doing some crazy stuff. Also use Euler a karras 30 steps with this model. Let me know if any of that helped. I use vpred now and be been sick for the past week so I forgot a lot about pony already.
One of my favs
I used this in webui forge, and it didn't recognise many characters. But illustrious version works fine.
I'm in love with this model!!
But it doesn't work on CivitAI anymore. Is there any replacement? The other models I've tried aren't as brilliant...
Yeah I relied on this a lot. Im gonna have to basically re-learn Pony at this point.
Why it's no longer available to create?
Due to website limitations, I can only have two models for online generation. I've assigned the permissions of this model to another one.
@WAI0731 I used it for Persona characters. But now that it's no longer selectable, I can no longer continue
@yamatoroger64625 I’m really sorry about this. It’s due to the website’s limitations, not because I don’t want to keep it available for use.
Because this model has earned 20M generations ($250K) for CivitAI.com, I wonder if the website will make an exception and allow you to host 3 models?
@WAI0731 Damn those limitations
@rabbit_fox I've asked similar questions before, and the answer was that it's not allowed because I would also receive BUZZ income, which is unfair to others. The issue is that, at this stage, BUZZ can't be converted into cash. For me, it has no use at all.
@WAI0731 At least it's on PixAI
The loss of this model for the web generator is a huge blow. I used it for basically everything I did on pony.
There's no way 😭 no longer available to generate on website? This is the most basic model in this site 😞
This is the main model I used on CivitAI. Sucks we have to switch sites after all the time and money spent. Can't even remix or img2img old images.
It's back 😭😭🔥💯
@Der_Machtige_Ritter We are so back!
This is an epic model;
I'm not sure what you did to get it back but thank you thank you thank you!
I mostly use illustrious and don't use pony. His illustrious model is really good, is his pony model the best for pony?
a
I have problems using poses, it does not work.
Which one are you using? try this https://huggingface.co/xinsir/controlnet-union-sdxl-1.0
@WAI0731 I use poses like legs up, split or wariza but they don't work for me.
@senjuitadori465169 This is a flaw in the pony model. If adding weights doesn’t work, then it basically means we need to load LoRA.
这个模型对SDXL控制网的线稿控制兼容非常好（我使用union），图生图时风格转绘很强，画错的地方修复效率也高
This model is very compatible with the line art control of the SDXL control network (I use union), and good at style transfer (I make 3D render to anime style). It also has high efficiency in repairing errors in drawing
My favorite
The best one👍
VERY COOL!
Amazing
This is amazing.
Nice
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.