@4044 关注了，哈哈

@4044 No problem at all. Thank you for responding.

@4044 44兄，现在这个V4和之前一上来你发布的v4一样吗

@4044 谢谢兄弟。

@4044 v2的色彩会好很多诶，v4饱和度明显降了一截，我试试

https://civitai.com/models/43227/oc

@polofish 感谢

要下载一个inpaint专用的底模来一起用

@4044 好的好的，感谢

鬍子比較難計算

#男女平等

@4044 真是一个悲伤的故事

Something like "girl,(looking away:1.1), from side" should work https://i.imgur.com/txYmg6b.png

@4044 一开始用16yo，基本看不出变化，一直试到12yo，可以看到面容有变年轻，但是和提示词的要求还是有差距

@izitn 加个bishoujo试试

Lmao

瑟瑟是第一生产力🤣

艺术有框架吗，没框架！不穿衣服也可以体现构图的艺术美感，想想文艺复兴。不要小看lsp啊喂

@xuecuichengvip314 艺术确实不应该拘泥于形式！但是如果不够美观众也可以批评哈哈哈，我就是文艺复兴风格的后现代艺术家！

这叫对人体的热情！

怎么画

@LeonTheLord 看c站看着看着就打开了ph

@Shaking 搞艺术嘛，需要灵感

@Shaking 冲完就可以安心炼丹了

用light room直接就可以，干掉痣很简单。你在这儿还要去找原生训练模型的问题，麻烦的很。

@4044 it still generated a lot of moles, i use this model for generating men pics, it will generate some skin spots(skin imperfection) which makes the pic realistic and i'm totally okay with that, the pics are just perfect before upscaling, but when using multidiffusion or ultimate sd upscale to upscale those pics, those skin spots(skin imperfection) will transfer to black moles. i found a workaround is that i put skin spots, acnes, skin blemishes, age spot in adetailer extension negative prompt, this will make the face clean, but body part will still have moles after upscaling.

@4044 是的，positive prompt不写，negative prompt只用easynegative，大佬方不方便说一下图生图你用的采样方法，步数，cfg，denoise strength，我发现用dpm++ sde karras会更容易产生很多痣

@4044 谢谢大佬

有痣或雀斑是比較自然，猜測原因是這個模型訓練是用大量張員瑛照片練成，因為特徵都保留著

长痣就更瑟了

长痣才正啊！！！

我知道你不爱看，你发给我，我爱看

哈哈哈

hhhh我一个正经刚毕业女大，在公司搞出来裸体，一时之间不知道是保存还是关电脑哈哈哈哈哈

适应一下就好了，不喜欢的都发给我，我给你穿衣服的

反向tag加上nsfw就好了

呵呵

我不信，你发给我看看

@afeng997 https://mega.nz/file/FykW0KaL#jsbh5jDNyS0raEckNkC24KOQs93PKzr2aR5j-R1xs0s

@huangsheng https://mega.nz/file/FykW0KaL#jsbh5jDNyS0raEckNkC24KOQs93PKzr2aR5j-R1xs0s

@Shaking 哥们你真发啊

女高觉得很常见

。。。

记得加sfw，或者负面提示加nsfw

@Una0915在公司记得提示词加 sfw，意思是safe for work，或者负面提示加nsfw， 

@jiaoling 你头像有点帅

怎么能证明你是正经女大学生？

@inurDaydream 确实，加了就好了

@inurDaydream 俺之前加啦，但还是不行。请教了同事，还算正常现象

(bad hands)

加texture inversion

https://huggingface.co/naonovn/Lora/tree/main

@LittleBoy1718 大佬请问<lora:LoRA_train:0.4>在哪找啊，你们是怎么在huggingface上找到这些模型的啊。😂

感觉是 SD 版本升级后，LORA名称显示的是原作者训练时随便写的名称，而不是实际的名称

@ali_ali 嗯嗯，应该是私模，谢谢告知

I'm glad to hear that you find the model amazing! As for the pixelated images you are getting when trying to run upscaling through "Send to Extras", there could be a few reasons why this is happening.

One possibility is that the input images you are trying to upscale may not be of high enough quality or resolution for the model to produce a high-quality result. In this case, you may want to try using higher quality images or increasing the resolution of your input images before running them through the model.

Another possibility is that there may be a technical issue with the upscaling process. You may want to try troubleshooting the issue by checking for updates to the software or reaching out to the support team for assistance.

It's also worth noting that upscaling low-resolution images to ultra-high resolution may not always produce perfect results, as the model is essentially "guessing" at the missing information. Therefore, it's important to manage your expectations and understand that the final result may not always be perfect.

I hope this helps!

哈哈哈 现在怎么看到以前小眼睛那种模型

真的吗

FilmGirl 胶片风 Film Grain LoHA & LoRA - Provia / FilmGX2 LoHA | Stable Diffusion LyCORIS | Civitai

你可以试一下这个

me to

Select “manage” in Model and you can find a place to load the file

去抖音或者B站搜索有详细教程，

【AI绘画】全部模型种类总结 / 使用方法 / 简易训练指导——魔导绪论 - 哔哩哔哩 (bilibili.com)

@kerrygoldenpig Thanks for the link. But that's for Stable Diffusion, not the program I am using.

@zijieliu2001124 I was able to load the model. How about LoRa? Where can I browse and try out? Lastly, I can generate images, but how do I train the AI?

https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors

You're probably missing some of the resources the author is using. If you want to replicate these, make sure you have Dynamic Thresholding and ADetailer extensions installed, badhandv4 and ng_deepnegative_v1_75t negative embeddings, as well as 4x-UltraSharp upscaler model. :)

@airi Except for the 4x-UltraSharp upscaler model, I have used everything else, but it still doesn't work.

@daisr0211181 Could you share a result with embedded metadata?

@airi I reinstalled the model and the problem is solved! Thanks a lot for your help！

@4044 thank you 😀😀😀