Sorry for the late upload, and also for my speech. I explained in quick and simple terms the functions of each tab in the Automatic1111 WebUI My Discord group: / discord
Thanks for the video. I did not expect to learn to much since I've been using SD for a month now. And it is true not to much of this helped me. Except it pointed out a few extensions to me that I'm not aware of. lol I just got into extensions for the first time yesterday. You kind of did not really explain any of those extensions you use. That's fine since this was a beginners video. But I know the name of those extensions from your video. I'm going to go research them now to see if I want them. So overall this video still helped.
The Tagger is used to get the captions from image dataset in dannbooru style tags for model training, openoutpaint is used for extend the images and the clip extension.. well.. i dont know, i guess it got installed with the tagger, it's used to see what version of clip captioning the model uses. Clip captioning is a way to get the prompt of an image. These concepts are still advanced and that's why i avoided explaining them in a beginner video.
Edit2: I think vae file can be put along side the model file cause in settings it says "it will check if it is there" right? I might be wrong. VAE is only available for version 4.0 I saw.
You can put the VAEs either in "stable-diffusion-webui/models/VAE" or in the folder where you put your models, guess it will be the same. Yeah, it's only available for Any4.0, but I guess you can use it alongside 4.5, it won't be much a difference. If you see that the image doesnt generate faces well or artifacs even with negative prompts, change the VAE to automatic or none. But if you use for example the Any4.0 Vae for let's say.. the default stable diffusion 1.5 or 2.x, then it would be a problem.
"Clip skip" seems to have significant effect on a "complexity"? I tried like "Clip skip = 1" and "4" - and image seems to be more refined and it made new version of the image, although the seed is the same
CLIP is an advanced neural network that converts prompts into numberical representation made of many layers, but you can stop early, at 1 or 4. But well, it's all about personal preference, but if you said it made better images, then maybe it did. CLIP is used to produce an image basically. But I don't know if I can say it's the "complexity". Complexity can be the resolution, steps, the upscaling and so on. I can guess that "clip" maybe.. I guess? follows your prompt better.
Thanks for making these videos. Really appreciate it. :) Subbed.
worth adding segments to the video for new viewers coming back
Thanks for the video. I did not expect to learn to much since I've been using SD for a month now. And it is true not to much of this helped me. Except it pointed out a few extensions to me that I'm not aware of. lol I just got into extensions for the first time yesterday. You kind of did not really explain any of those extensions you use. That's fine since this was a beginners video. But I know the name of those extensions from your video. I'm going to go research them now to see if I want them. So overall this video still helped.
The Tagger is used to get the captions from image dataset in dannbooru style tags for model training, openoutpaint is used for extend the images and the clip extension.. well.. i dont know, i guess it got installed with the tagger, it's used to see what version of clip captioning the model uses. Clip captioning is a way to get the prompt of an image. These concepts are still advanced and that's why i avoided explaining them in a beginner video.
Edit2: I think vae file can be put along side the model file cause in settings it says "it will check if it is there" right? I might be wrong. VAE is only available for version 4.0 I saw.
You can put the VAEs either in "stable-diffusion-webui/models/VAE" or in the folder where you put your models, guess it will be the same. Yeah, it's only available for Any4.0, but I guess you can use it alongside 4.5, it won't be much a difference. If you see that the image doesnt generate faces well or artifacs even with negative prompts, change the VAE to automatic or none. But if you use for example the Any4.0 Vae for let's say.. the default stable diffusion 1.5 or 2.x, then it would be a problem.
@@grendarAI Oh, I see. I tried both 4.0 and 4.5 - if "feels" that 4.0 is a bit more consistent with those vae file, but I am not sure.
"Clip skip" seems to have significant effect on a "complexity"? I tried like "Clip skip = 1" and "4" - and image seems to be more refined and it made new version of the image, although the seed is the same
CLIP is an advanced neural network that converts prompts into numberical representation made of many layers, but you can stop early, at 1 or 4. But well, it's all about personal preference, but if you said it made better images, then maybe it did. CLIP is used to produce an image basically. But I don't know if I can say it's the "complexity". Complexity can be the resolution, steps, the upscaling and so on. I can guess that "clip" maybe.. I guess? follows your prompt better.
@@grendarAI ya, better would say - it different "clip" create different variation
Just a question - what is the prompt for the intro photo?
files.catbox.moe/xjjk9o.png there you go mate, download it and drag the file in PNG info tab and move it to txt2img
@@grendarAI thank you. Will try.
Any website for ai anime mages and the prompt sharing ? 😅
general.aibooru.online/
Too many weebs in the stable diffusion world. You guys are ruining the place. I'm SICK of anime! FFS, enough already.
@@peterbelanger4094 We are sorry for that
@@grendarAI It shows how non creative y'all are. Just wankin off to anime girls all the time. That's not art.