我一直想在Playground v2中复现我在SD1.5所使用的二次元画风，但毫无疑问，我失败了。

这个lora仅是一个试水作，目的是为了测试该模型对眼睛特征的学习，就结果来看，眼睛似乎学得还可以，这个是唯一的好消息。

值得注意的是，该lora并没有对星野爱 / Hoshino Ai 的官方服装进行学习，甚至我的训练集中就混进了很多没穿衣服的素材，因此如果您在测试lora的过程中发现会随机出现裸体，请不要惊讶，这不是您的问题，解决这个问题的方法也很简单，只需要加入一些服装的标签就可以。

在SDXL1.0中，我并不能找到一个支持nsfw的二次元大模型，而Playground v2的出现让我看到了希望，我大概还会不断尝试，祝我好运。

I've been trying to replicate the 2D art style I used in SD1.5 in Playground v2, but undoubtedly, I have failed.

This LORA is just a test run aimed at evaluating the model's ability to learn eye features, and in terms of results, it seems that the eyes have been learned fairly well. That's the only good news.

It's worth noting that this LORA hasn't been trained on official costumes of Hoshino Ai or any specific character. In fact, my training dataset included a lot of unclothed materials, so if you encounter random nudity while testing LORA, please don't be surprised. It's not an issue on your end, and the solution is quite simple—just include clothing labels in the training data.

In SDXL1.0, I couldn't find a large-scale 2D model that supported NSFW content, but the introduction of Playground v2 has given me hope. I will continue to experiment and hopefully, luck will be on my side.

这是基于Playground v2训练的Lora.

触发词为：hoshino ai (oshi no ko)

建议使用的分辨率：768*1024

基于最新的模型:playgroundai/playground-v2-1024px-aesthetic · Hugging Face

我训练和测试的模型是其他人修改过的p16版本：lrzjason/playground-v2-1024px-aesthetic-fp16 at main (huggingface.co)

如何使用：像SDXL1.0那样正常使用就行，底模需要换成我上面说的fp16版本

您可以使用：

https://github.com/AUTOMATIC1111/stable-diffusion-webui

或者：

https://github.com/comfyanonymous/ComfyUI

在SD-webui的使用

请把lrzjason/playground-v2-1024px-aesthetic-fp16 at main (huggingface.co)

../stable-diffusion-webui/models/Stable-diffusion 路径下

然后把lora放到

../stable-diffusion-webui/models/Lora 路径下

在ComfyUi的使用

请把lrzjason/playground-v2-1024px-aesthetic-fp16 at main (huggingface.co)

../ComfyUI/models/checkpoints

然后把lora放到

../ComfyUI/models/loras路径下

This is Lora trained on Playground v2.

Trigger word: hoshino ai (oshi no ko)

Recommended resolution: 768*1024

Based on the latest model: :playgroundai/playground-v2-1024px-aesthetic · Hugging Face

The model I trained and tested is a modified version by someone else:rzjason/playground-v2-1024px-aesthetic-fp16 at main (huggingface.co)

How to use: Just use it like SDXL1.0, but replace the base model with the fp16 version I mentioned above.

You can use:

https://github.com/AUTOMATIC1111/stable-diffusion-webui

or

https://github.com/comfyanonymous/ComfyUI

For SD-webui usage:

Place lrzjason/playground-v2-1024px-aesthetic-fp16 at main (huggingface.co) under the "../stable-diffusion-webui/models/Stable-diffusion" path.

Then place Lora under the "../stable-diffusion-webui/models/Lora" path.

For ComfyUi usage:

Place lrzjason/playground-v2-1024px-aesthetic-fp16 at main (huggingface.co) under the "../ComfyUI/models/checkpoints" path.

Then place Lora under the "../ComfyUI/models/loras" path.