使用comfyui这种支持V预测的话，与EPS相比，V更好些？

所以esp v1.1相比v1.0提升了什么 没看出区别

作者你好，我现在用的EPS1.0，我想问一下1.1对比1.0增加了什么或修改了什么吗？可以告诉我吗？

I really like the skin texture you can get in this model, the characters have a nice look and lining to them, pretty decent in bg drawing as well, overall I would definetly recommend trying it out.

无提示词渲染的效果真的很棒!!!, 超级可爱!!!

但是对Lora的适配度不高, 很大部分Lora权重稍微大点(比如超过0.5)就会有蓝绿色伪影

It's not just the skin, it's also the amazing lighting that brings out the undertones and overtones of the skin as well. The dark lighting environments this model produces work wonders for Wan and Grok animations because they come out a lot smoother than most other image models when Wan and Grok interpret the "depth" of the image.

很棒的皮肤光泽感

any suggestion to inpainting in this model? it looks great but when inpainting even at low denoise it´s messy.

用了快一周了，很喜欢这个模型的画风和质感，就是手脚有时还是有点问题，而且我发现带张嘴的tag很容易出现彩斑不太清楚为啥，调高步数会好一些，期待新版本

Skin texture and lighting is very refreshing, but I wish the same was for hair. It's really hard to prompt for textured hair with strands visible, instead mostly i get "liquid" or simplified-vague-anime-hair texture.

On the flip side, I really like how it does plump\chubby bodies, and huge breasts + saggy breasts gives actually pretty, natural-looking breasts; while with models like wainsfw or etc you either get inflated balloons or SAAAG and have to carefully tweak loras to get close to where you want, this model has good looking proportional realism baked in it though.

It would also be a little nicer if plump\chubby tags automatically gave thick thighs, often you have to prompt them separately, but that influences the composition in ways you don't always expect! The model is also better than others at not giving too hourglass-like waists when you say plump\chubby, but there's still a little room for even more naturally proportioned waists.

P.S. i'm not generating ugly bbw girls, i'm using plump\chubby tags for nice-looking mature realism. Plump (not fat) girl supremacy!

The overall look of the model already shines through with its balanced textures and lighting choices. Skin comes through with a clear, refreshing presence, and the subtle structure across the body reads as intentional rather than exaggerated. That sense of grounded design is part of what gives the model its recognizable charm, and it would be wonderful to see that core direction preserved.

There is also something genuinely appealing about how the current proportions lean toward natural softness without pushing into extremes. This gentle shaping gives mature realism without drifting into distortion, and that restraint feels like a defining strength. Models that chase overly amplified forms often lose harmony in the silhouette, but this one maintains a consistent, grounded proportional logic that sets it apart.

For those who prefer more pronounced size adjustments, tag-based control already offers a flexible pathway. Allowing individuals to apply their own preferred modifiers keeps the foundation stable while still giving ample room for personal customization. This helps the model stay true to its identity while letting enthusiasts tailor their own results through additional prompts.

The existing waist–hip balance, the tasteful rounding, and the soft but coherent body lines all contribute to a signature aesthetic. Preserving that baseline ensures the model continues to offer realism that feels intentional rather than forced. Extra volume can always be added through specific prompts, but the core template benefits from remaining centered and proportionally consistent.

Keeping this original philosophy intact highlights the thoughtful structure already present, while those seeking more extreme fullness can freely build on top with their chosen tags. This way, the model remains recognizable, flexible, and expressive without losing the clarity that makes it stand out.

The issues with hands and feet in 1.1 seem to have been addressed in 1.2, but overall I am finding that the actual image composition is significantly worse. Heads are out of frame in ~90% of images and adding composition tags like full body or whatever as recommended making very little difference to that, the majority of the time it's ignored and head is still cut off.

I also generate in large batches with a lot of wildcards so adding composition tags is also really impractical for me. Overall I was getting better images out of 1.1 than 1.2 with the same prompts - having to throw away ~20% of them due to bad hands/feet or bad skin color is still preferable to having to throw away ~80% of them because they're all close ups with head out of frame. So will be going back to 1.1 for now. Hopefully this can be addressed with another revision.

v1.2esp好像相对于v1.1全面劣化了 不管是提示词的遵循程度还是整体画面的控制力，甚至1.2在laten放大的步骤会直接出现肢体畸形很恐怖。v1.2好像是失败的模型？同样的设置1.1完全更好，这是什么道理。总之总体看戏来好像1.1出来的图成功率更高？

Everything else in this model is good, but the face control is too strong, and almost all the characters have the same face shape.

A new version is out! Hooray! I was using V-pred 1.1 before.

It's been a while since I updated, so I'm giving this a try: version 1.25.

新版本出來了～灑花。之前都在用V-pred1.1

久違的來更新一下試試1.25

这是我最常用的2种型号之一。 感谢您的努力。