Sleep with open eyes - 目を開けて眠る - 睁眼睡觉 - SD1.5

https://civitai.com/models/584231/sleep-with-open-eyes-sd15v11?modelVersionId=654038

My English is not good, so I used ChatGPT to translate this introduction.If you can use the Pony model, I recommend using the Pony version of the LoRA. Click hereI am a beginner at training LoRA models, so this LoRA model is a test version.This LoRA is not only suitable for sleeping scenarios; it is a facial expression concept that can be used for any scene you imagine. It mainly affects the character's facial features, making the eyebrows and eyes appear particularly relaxed, weak, blank, and lifeless.I have to complain about one thing. This LoRA model was initially created using the PonyXL model and could be easily used. However, I encountered many issues with the SD1.5 version of the LoRA model due to the limitations of SD1.5.Compared to the Pony version of the LoRA, SD1.5 lacks many details and has more problems, including poor facial generation for inverted and side-view characters, low native resolution leading to missing eye details, insufficient understanding of prompt words (e.g., (half-closed eye - narrowed eyes)), and issues where eyes are forced to fully close when using (sleep - sleeping) or (closed eyes), (looking away) is ineffective, and SD1.5 often makes characters stare at me.Trigger words (non-sleeping): HJG, (half-closed eye:1.7)If you use anime character LoRA models, please use this more effective trigger word: HJG, (half-closed eye:1.3)LoRA weight recommendation: 1.01~1.2Negative prompt words recommendation: (blush:1.2), bags under eyes, lips, (eye_contact:1.2), short fingersIf you have the LoRA Block Weight plugin, these are the parameters suitable for this LoRA: “lbw=1,1,1,1,1,0,0,0,1,1,1,0,0,0,0,0,0”It is recommended to always enable "ADetailer" to assist in fixing facial details.It is recommended to always enable the "Hires. fix" high-resolution repair function to ensure there are enough pixels to carve eye details.There is a common issue with this SD1.5 version LoRA model. If you need to generate (hand pov) type images, you can ignore this issue. However, if you don't need the character to always interact with hands from outside the frame, please add “(pov hands:1.4)” to the negative prompt words.LoRA needs to ensure the eyes are visible, so do not use “sleeping” to make the character lie somewhere. Instead, use “lying” to guide the character to lie down, and use some features of ControlNet to control the character's pose for better results. Sometimes you might get an image where the eye color pollutes the entire white of the eye. The reasons for this are varied, so try reducing the weight of (half-closed eye) and <lora:HJG - SD1.5 v0.2-000006>.Sometimes, you may not be able to get the character to open their eyes. Adding an eye color to the prompt, such as "red eyes," has a small chance of improving the issue (I can't guarantee it will work >_<).SD1.5 is very poor at generating faces for characters that are inverted or looking sideways. Please use ControlNet to avoid this.I recommend using another author's LoRA [eye opening (eye size) control] to fine-tune the control of eye openness.————————Currently, I have found that this LoRA has varying compatibility with different styles of large models, and the prompt words need some fine-tuning. I will add an XYZ chart later to help everyone understand how to adjust the weights of the LoRA itself, (lying), and (half-closed eye). Work in progress... ... ...These descriptions are just part of my understanding of this LoRA model. Feel free to explore more ways to use it boldly and share with everyone! ❤️