Thanks for the FP32 pruned version of V3! Looks amazing! 🔥

The V3 version has better texture and more realistic visuals, but it is easy to generate European and American facial shapes when using Asian LORA, making it difficult to generate the original appearance of LORA. Therefore, the V2 version is more suitable for Asians.

請教大佬這個主MOD跟另一個容量較細的LORA分別在於？

所以 这到底是不是一个动漫的模型？为什么我的用起来是动漫的？

哥，一个月更新1次就好了~

这款模型的inpainting是我用过效果最好的，只不过一直用的是v1（看介绍v2的手不行）,一眨眼现在已经v4了，期待新的inpainting能带来惊喜。

弄个小红书风格的呗，人瘦腿又白又长的那种，

总感觉V4出人物图不如V3，不加LORA出的都是很丑的欧美人物，加LORA也几乎没什么用

Any possibility to use this through Hugging Face and Google Colab? Looks very interesting.

anyone know how to trigger small boobs? because i keep try using small boobs or small tits, still got big boobs. In inpaint.

用FP16 做inpainting的話效果會跟v4 pinpointing 的差很多嗎

这是我用过最好的修复模型！太赞了

How can I use this model? Is it on hugging face?

nipples arenot clearly formed,anything helpful would be thanked!

Pro got way better but it’s over 10 gb of size that really could’ve stressed it but, use the On-site Generation for Purposes.

Anyone know what "Error message: 'encoder.conv_in.weight'" means? I get this when trying to import all 3 versions of this new model. Never had that before.

Thanks for the fixed update. Work well. I am also testing it with the LCM Lora and sampler in A1111 with pretty good results. It does degrade the output a bit but it outputs w/ Hires in like 1 second! (rtx 4090).

loving the fantasy elements!!

This checkpoint should really be the gold standard. It does everything well, not just NSFW stuff! Thanks for sharing your work.

I can't reproduce any image from the preview by copying the exact settings

Wait, version confusion: Is this the fix to the previous v4 model? So the v5 is still the most current?

Is the Asian V1, in the original Pornmaster model or is this version missing?

Im using ComfyUI on Runpods, how do I install this checkpoint using the prompt. The standard -O Name.safetensors is not working

Hello,

How to use this base model with fooocus ?

In "Base Model (SDXL only)" pornmaster pro don't load, is it used as a refiner ?

And if so what base model i should use ?

Thx

Excelente option!

thank you and v8 inpaint come on~~~~!!!!

I've never been able to get these models to produce anything like the examples. It seems it's made to be used with LoRAs only.

I mostly have experience with Pony Realism. Vs. Pony Realism I really like the style of PornMaster BUT it's also a lot more challenging to use. I am using XYZ plot a lot to see what the right sweetspot is to not turn generations into something incomprehensible. Many of the lora aren't that compatible with each other and adding more than 2 lora has a high risk of totally destroying your generations. Upscaling also destroys images unless you do tiled or inpainting. I find this works best when you go for low accuracy generations and then focus on inpainting with the specific lora that would be most helpful. I really love the style of the pornpolaroid lora, but it has very low compatibility with other lora. I wish I could just use these LORA in PonyRealism

are these models finetuned at 512, 768, or 1024?

@iamddtla

What happened with v10?

Only the pruned model is available for download. Can we get the full model somewhere?

Please SDXL

Your description states that there are two higher grade models, k.e. like 4GB and the unpruned version. Do you still publish them somewhere?

Flux someday?

看起来很不错的样子

V9. A very decent realistic model. I would especially like to note the good correct anatomy, in my tests there were almost no deformities of the limbs. Despite the name, it does not aim to reveal boobs to everyone and also works well with male characters and a wide range of themes, whether they are animals, cars or landscapes. Excellent flexibility and responsiveness. Great job!

老哥，我用1.9GB亚洲v3版本模型来训练SDXL的LORA提示safetensors_rust.SafetensorError: Error while deserializing header: MetadataIncompleteBuffer错误，是这个模型不能用来训练lora吗？

INPAINT我研究很久了 从brushnet到FLUX 肤色、光影、油腻感要完全对上真的很难

Probably one of the easiest to use models and the results are solid even with less specific prompts/ weight sliders, you can create decent generations.

But of course if you want to generate epic renders, you need to combine with lora's, upscale, inpaint ect. :)

Anyways, this is one of my favorites and also the no. 1 source when it comes to asian models

wow you got your pony model's humans to not look like 3 flavors of Bratz Dolls. nice jorb

I love the new Pony models. Would you be able to give us more information on the captioning strategy that you used? Booru tags don't seem to work very well compared to base Pony so it'd be good to know what the captions looked like for training.

It would be awesome if you could give us a list of the most commonly used tags and phrases in the training captions - that would really help with learning how to prompt this model

its a shame your 10.1 version was removed from civitai generation services recently

i would use the Unreal checkpoint version with TensorRT but it seems to not have an Onxx file in it, so i can't use it as well as i want to, i have tried over 10 times to export it, but fails and crashes my webui console.

any advice/recommended training parameters on training lora with your model/checkpoints? my images are always saturated...

Why did you use score tags in the xl version?

Phenomenal model!

wowwwwww SDXL!!

Best settings for pony-v1-inpaint? in forge?

Is this model good to train a LoRA of a real person?

Will you make a guide on train lora with your model? I tried training many loras with it but none came out good looking. Even though you said it can do many races, your 10.1 always produce chinese faces

Help its not working on my Comfyui. i got pixelated images and very bad image. i use ur setting DPM++ 3M SDE steps 25 cfg 6 and also LCM steps 8 cfg 1.
I use SDXL Vae. why is that?

Hello, I like a lot of models! but someone could explain to me how to edit the yaml file to create a Lora with this model on Fluxgym please? I do not understand ...

конкретно эта модель генерирует реалистичней всех моделей, которые можно найти в сети, с нетерпением жду новых версий

Would love to see this be useable on civitai

I don't know what kind of magic is happening here but this is the ONLY checkpoint where I can crank up my LoRa strength even more without getting any distortion! Also LoRas seem to work really great and have a stronger resemblance than with other checkpoints!

5 more hours

best model indeed, i wounder if there is any planning for the slow as hell flux, anyway thanks a lot

someone explain Illustrious to me. Pony but way uglier? why??

想請問noob-V0.2有增加什麼嗎?因為昨天剛試用V0.1而已。下載的話還滿久的，謝謝

Great work! Could you make an article how you processed images and tagging?

Thank you for this model. Will SDXL version become available to generate on CivitAI again?

didn't try it but I doubt it's better than Lustify

How do you make models that doesn't have asian faces?

Hello! I tried to use SDXL-V1.1-VAE-inpainting in A1111. But I keep getting this error:

modules.devices.NansException: A tensor with NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.

Is there a right way to fix this?

impressive realism, but horrible prompt adherence