thiros

/

YuzuLemonTea

like 186

Hi, thank you for downloading my merge models.
I'm gonna add prompt of 1st image of 3girls on a street in model card, please check it.

Thank you for your help. I'll use it well.

Hi thiros,

Thank you for sharing the model!

I am noob in this field. I tried the query "YuzuLemonTea", but I didn't find the toutorial.

I am wondering if you could give some pointers/hints on how to use these model files?

Hi, thank you for being interested in my model.
To use models, simply download .safetensor files and put these in stable-diffusion-webui/models/Stable-diffusion folder.
You will be able to choose models in webui gui.

If you have no clue what the text above means,
these models are for local AI image generation Stable Diffusion WebUI.
https://github.com/AUTOMATIC1111/stable-diffusion-webui

Installation guide is here:
https://rentry.org/voldy

There are also some communities if you search "Stable Diffusion" on google.

Hi, thiros,
Thanks for sharing the model and the relevant tutorial, it is useful!

I have installed the WebUI with Stable Diffusion and also put the .safetensor files into stable-diffusion-webui/models/Stable-diffusion folder.
Following your recommendation setting (Including the same prompt and negative prompt ), I can generate a complete image.

But I also meet a problem in that the generated image seems to lose some color, which results that I cannot repeat your example. It made me very confused.
I list the comparison image pairs(the in-generating image, and generated image) in this comment, you can find the in-generating image performs more light and contrastive than the generated image.

I have no idea to solve this flaw issue, could you give me some suggestions or hint to overcome this problem? I will so appreciate it.

Hi, Jadhrkin. Thank you for using my merge model.

It seems the issue occurred because of VAE. VAE plays big part of contrast and coloring.
As mentioned in model card, I usually use vae-ft-mse-840000-ema-pruned or kl-f8-anime2(depends on my feeling).
If you use original NAI vae and haven't try others, follow the instruction below.

1. Download vae-ft-mse-840000-ema-pruned.safetensors and kl-f8-anime2.ckpt from links.
   https://huggingface.co/stabilityai/sd-vae-ft-mse-original/tree/main
   https://huggingface.co/hakurei/waifu-diffusion-v1-4/tree/main/vae

2. Rename them to vae-ft-mse-840000-ema-pruned.vae.pt and kl-f8-anime2.vae.pt, and put them into /stable-diffusion-webui/models/Stable-diffusion.

3. Start WebUI, go to setting tab, select "Stable Diffusion" from left menu, choose vae from drop-down(vae-ft-mse-840000-ema-pruned.vae.pt or kl-f8-anime2.vae.pt whichever you like) and click "apply setting."

vae-ft-mse-840000-ema-pruned.vae.pt and kl-f8-anime2.vae.pt have similar coloring but bit different, so find out which you prefer.
Hope this would help

Thanks a lot for your prompt help, according to your advice I have solved this problem!
The newly generated image is as follows:

Finally, could you recommend some prompt words tutorials or relevant introductions for how to set the desired prompt words？THANKS!

I write prompt in following order;
art style(realistic, 3d render, illustration, anime painting...), subjective(1girl, 1boy...), clothes(wearing ~), verb(walk, talk, sit...), adverb(alluring, cheerful, happy...), facial(smile, angry, crying, open mouth), location(street, school, train, forest, city name)

but it's just my style(I change sometimes) and is not the best and only way.
I see some start from place and others write sentences.
prompting is flexible and there are many way to write.

This web page have many example. I guess this will help you.
https://zele.st/NovelAI/