Veriwarp | 逆幻 (Illustrious)

大家好，我是ekll。

Hello everyone, this is ekll.

欢迎来到Veriwarp(逆幻)模型，这是一个基于rouwei的融合模型，现在请放弃您习惯的一切“常识”。

Welcome to the Veriwarp (逆幻) model, a merge model based on rouwei. Now, please abandon all the 'common sense' you're used to.

如果您喜欢该模型还请点赞并量力而为的支持我一些buzz以供在Bid中支付在线生成的费用。

If you like this model, please give it a 'like' and consider supporting me with some Buzz (if you are able) to cover the online generation fees in 'Bid'.

我的推特:@ekll

My Twitter: @ekll

Join Discord

这个名字是由

Veri-:拉丁语 "Veritas"，意为“真理、真实”。

-warp:英语 "Warp"，物理/艺术: "Warping reality" (扭曲现实)，代表着变形、扭曲、重塑、改变。在Photoshop等工具中，“Warp Tool”就是用来进行自由变形的。

The name is composed of:

Veri-: From the Latin word Veritas, meaning "truth" or "reality."

-warp: From the English word "Warp." In the context of physics and art, it refers to "warping reality" and represents concepts like transformation, distortion, reshaping, and alteration. For instance, the "Warp Tool" in software like Photoshop is used for free-form transformations.

目前支持的年龄tag

女性：mesugaki、bishoujo、lady

男性：shota、bishounen、gentleman、geezer

1girl、1boy已废弃，请您用以上tag代替1girl、1boy。

Currently supported age tags

Female: mesugaki, bishoujo, lady

Male: shota, bishounen, gentleman, geezer

1girl and 1boy are now deprecated.Please use the tags above instead.

正面提示词中不需要质量tag

负面提示词:"masterpiece,best quality,worst quality,low quality,watermark"。

是的，您没有看错，这并不是写错了，还请在负面提示词中填写以上tag。

Quality tags are not required in the positive prompt.

Negative Prompt: "masterpiece, best quality, worst quality, low quality, watermark".

Yes, you read that right, it's not a typo.Please use the exact tags above in your negative prompt.

关于外貌穿着类提示词：请尽量不要同时使用含义重复的的tag。例如dress,long dress,blue dress,frilled dress请直接写成blue frilled long dress。

For prompts about appearance and clothing: Try to avoid using tags with redundant meanings simultaneously. For example, instead of writing dress, long dress, blue dress, frilled dress, combine them directly into blue frilled long dress.

关于动作类提示词：可以适当使用自然语言来代替tag描述角色动作，通常会比直接使用tag略微准确。

For prompts about actions: You can use natural language to describe a character's actions instead of just tags.This can often be slightly more accurate.

生图分辨率

分辨率应介于1024*1024（总像素1,048,576）-1536*1536（总像素2,359,296）之间的任意尺寸。

低于1024*1024（总像素1,048,576）通常会使图像变差，高于1536*1536（总像素2,359,296）不会让图像产生质变。

个人经常使用960*1344/1096*1536。还请您根据图片复杂度进行调整，并不是越大越好。

Generation Resolution

The resolution should be any size between 1024x1024 (1,048,576 total pixels) and 1536x1536 (2,359,296 total pixels).

Using a resolution below 1024x1024 (1,048,576 total pixels) will generally degrade image quality, while going above 1536x1536 (2,359,296 total pixels) does not yield a significant qualitative improvement.

I personally often use 960x1344 or 1096x1536.Please adjust based on the image's complexity, as bigger isn't always better.

采样器与调度器

er sde/kl optimal，如果您不清楚这是什么还请进行尝试！

当然其他采样器与调度器也可以正常使用。

Sampler and Scheduler

er sde/kl optimal. If you're not sure what this is, please give it a try!

Of course, other samplers and schedulers will also work fine.

CFG：4~6都可以正常工作。

步数：20~40步是一个通常的设置。

CFG: A range of 4 to 6 works well.

Steps: 20-40 is a typical range.