Best LORA ever !!

When i saw your post on reddit this morning, i thought to myself, "well, this is gonna go viral". And it actually did 😆.

Anyway, great LoRA.

很好的罗拉，爱来自新的山坡♥

Fantastic LoRA idea, very useful for tweaking the art styles of an existing checkpoint or LoRA that you like!!

Feels like magic, great Lora!

Been looking for something like this for a while now! Thank you!

🤩

请问适合inpainting么

减少细节这一点实在是太厉害了

Holy shit this LORA works almost perfectly

i don't know what happened

but it's awesome!!!!!

Who doesn't want to have a "detail slider" in their prompt?

If the results are as good as pictures and comments suggest here and don't produce weird artifacts, I'll probably use this LORA even more frequently than epiNoiseoffset.

what model are you using？

it's bizarre how well this works, wonder how it was made

Are there any plans for 2.1 version?

I love this LoRA!

It doesn't seem to be block-merge-able, which is a bit sad.

Absolutely amazing how well this works. Improving all my previous works with this.

Excellent work!

by far most amazing Lora

Great Lora, I hope you make more!

请问这个LORA的训练思路是怎样的？很好奇，求江户川大佬指导~

May I ask what is the training approach for LORA? I am very curious.

Are there triggering words?

Wow, works exceptionally good... You can even set a low CFG, 1 or 2, and get nice details, contrast and color! This will be a staple for certain.

magical!

Good job!

Wow! amazing Lora!

my hero

确实很好用

但是这个lora权重太高或者迭代太多次会导致画面显得模糊或者线条太粗，用一次就好

Bro just changed the game forever

这个Lora确实能添加细节，但是写实人物的头发会变得线条感太强，不知后续能不能优化。

太牛逼了,这个功能性

Awesome LoRA! Any hints on how one would go about training a model like this? Thank you :)

好强！

Awesome！

这种lora是怎么练出来的啊，太厉害了！

Hi, have some Problems with Vlad Diffusion...

add_detail.safetensors: AttributeError

lora = load_lora(name, lora_on_disk.filename)

AttributeError: 'str' object has no attribute 'filename'

- some Ideas what to do?

大江湖战士出的LoRA就是好用

It changed the game's rule.

大佬的这个lora修手是真的吊！

amazing results! 👏

我靠，甚至无法想象这个原理....

incredible lora..

這個真的厲害

how do I get it to work?

This one is so cool!

What kind of black magic is this? however, it does what it does. I don't want to live without it anymore.

绝了！

Where is this model placed and how is it used?

好奇原理的各位，可以去 B 站看青龙圣者的「差异炼丹」那期视频。

I'm sad to inform you all that after testing this Lora many times on realistic model, it more often than not just ages people. This is because the added details are often duplicates, although very well placed duplicates, it duplicating wrinkles and smile lines often ages people so beware. This isn't an issue in anime Loras though cause those don't exist for the most part.

Did a video on this: https://youtu.be/M8-Y4t2T4bo

If anyone is interested check it out! Thanks for the LORA!!!

This is very cool, but how about SD2.1 version?

神中神

i use it on daily basis!!! THANK YOU !!!

抱歉請問該怎樣安裝……？

Why do I always get this ERROR (below) when I use this LoRA? Please point out for me the solution, I'm not a tech person so I don't understand how to solve it. Thanks! The ERROR as below (it always appears no matter when I use this LoRA solitarily or in combination with other ones. As long as I use this LoRA in my prompt, the ERROR shall appear):

"WARNING:tensorflow:6 out of the last 6 calls to <function Model.make_predict_function.<locals>.predict_function at 0x000001D92AF5E290> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for more details."

Please help me. Thanks!

I want to like this, i really do, but this model has to have a bad training base cause im getting replica renders back to back of the same thing no matte what i change on the negatives or positives.

Can you create this style for me? These are the images I collected for you https://mega.nz/folder/UDMkUSyQ#m1tqmdu9WueKEUY3RKBjYg

lora权重为啥是2?在我的认知里lora权重加起来不应该为1吗?

This LORA is amazing, I tested on every anime model I have and works flawless, I cannot say about realistic models but anime models are incredible, I recommend the use of the extension called additional networks, use with weight of 0,5 should be fine, thank you!

Amazing LORA, it also works well with inpainting clean up, i can use this to remove unwanted details

I am just very curious how to train a lora like this? What is the dataset you prepare for this?

It makes 2D become 2.5D

Hoping for a SDXL version! Best Lora ever created, without a doubt

This is a great Lora, but it seems to disable my Lycoris models when I use it, does anybody know why that might be??

And when I say 'disable' I mean it makes it so I literally cant even re-add any lycoris to my prompt (and have it function) even after turning this Lora off again, I have to reload the UI or change models to fix it! Very weird, any ideas??

EDIT- RESOLVED, SEE COMMENTS

must have

It's been my favorite lora now.

The only downside to this LoRa is its bias to pubic hair. Like, why???

Thanks for sharing these great Loras,they really help a lot.

The quintessential detail network for enhancing details. Thanks for this great LoRa!

I suggest that the category of this tool be changed from 'Concept' to 'Quality'.

https://tusiart.com/ https://www.liblib.art/

These two Chinese websites are stealing models from Civitai.

Please enable it for online image generation.

I can't understand the potential of this LORA but I'd like to try it.

Works really well, thanks!

How do you get the negative weights to work, all the systems I have tried ignore them completely

If we want to use this model to generate images for fee, could we get a license from you? If there is any chance we can get a license from you, please contact me.

Hey! Great work!
Your talent is just what my team's looking for in our project. How about we collaborate?

If you're interested, could you share your email or Telegram to discuss this?

Sorry guys, I'm really bad at this. What am I supposed to do after downloading this?

What a fantastic LORA, the improvement is truly remarkable . Thank you

Amazing results.

Which dataset did you use for training?

very cool depending on the model, and I've accidentally made horrors using this, too, lol

I'm new to this how do the downloaded lora's work.

Thanks.

This LORA makes miracles, OMG

Love it

Nice Lora, Thx.

probably Top 3 of the most useful SD 1.5 Loras ❤️

"Clip Skip" is what happens in stable diffusion ? I couldn't find anything with the same name. my English is not very good so I use translate.

Hi,

I am using StableSwarmUI. I have downloaded this model and tried to place in Stable-Diffusion and Lora folder. But i am getting error message like "Invalid operation: All available backends failed to load the model." while selecting this model in the selection. I am new to this. please help me how to add this model in StableSwarmUI.

This lora seem very memory hungry, just adding it increases the load in more than 5gb from what i could tell from my manager, so if you are struggling for memory like me this is almost impossible to use, sadly

I can't find this LoRa in list, when I try to generate images on this site. Why that can happen??

Really a good add to any generation.

It's a nifty tool to change the small details.

cant use in comfyui, create image quality is so low, why?

I've downloaded this LoRa but I don't see a trigger word. Is one necessary? is this applied in combination with other LoRas? Using Fooocus. Thanks!

Is there have new SDXL model to tweak detail?

seems it only work on SD1.5

xl版在哪

Good

works fine

无法用于comfyui，和xl

i need help.it doesn't work on forge

Nice, works fine