Bula

Bula

Jack-O-Lantern🎃
Jack-O-Lantern🎃
Rainbow Candy
Rainbow Candy
Pink Candy
Pink Candy
Green Candy
Green Candy
Red Candy
Red Candy
Blue Candy
Blue Candy
Hello!!, Thank you for using my model Please share your amazing art~ also check my other work Civitai : https://civitai.com/user/Bula
399
Followers
0
Following
1.4M
Runs
321
Downloads
7.1K
Likes
Latest
Most Liked
Guide how to prompt in anime model using Booru tags Halloween2024

Guide how to prompt in anime model using Booru tags Halloween2024

Halloween is Coming🎃🎃🎃WARNING! this article just for beginners. if you have mastered how to write prompts, I think you don't need to waste time reading this article.In this article I will write a guide how to prompt in anime modelsIntroAs you know, most anime models use caption WD tagger or booru tags. Why is it called booru? Because the model is trained with images taken from Danbooru or similar, they store large images with tags in them. From there, the anime model imitates how to tag each image to become a training dataset.What are booru tags?Tags are keywords used to describe images that you usually refer to as prompts.Booru tags use a method like describing something separated by commas. The more commas and descriptions, the more accurate the results will be.You can see the difference between booru tag prompting and natural language prompting below.How i prompt with Booru tags?So far, maybe you already understand how to use booru tags, yep right by separating by commas.But not all models work with random placement prompts, even if you have followed the comma-separated rule.In the new anime models, they have more structural placement, and there are even special cases in certain models, such as Pony models which I will explain later.So it's like this.First sentence is quality tags such as masterpiece, best quality, ultra detailed, 4k, and so on.Second sentence is a description of what you want in your art, such as 1girl, 1boy, group, etc.You must write it with details such as hair color, posture, and eyes, so that the results match what you want.3rd sentence is what you want the character to do, such as sitting, standing, running, or any other event you want.you can use your wild imagination.4th sentence is the time and place setting, this will usually affect the background and any events.How your prompt affects your imagesActually, the explanation is more complicated than this. if I explain about the tokens in each text and how the model processes text, it might be long and difficult to understand, considering this article is intended for beginners.This is a visualization of the image with colors. each represents the tags you input, which you can see below.How I use Pony model?Pony is a part of SDXL. due to the large dataset and unique training method, Pony turns into a different model.SDXL uses quality tags to improve the quality, unlike Pony which has a rank-score system."score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up"Score tags work like quality tags on the SDXL model. when you use the Pony model you don't have to use SDXL quality tags such as masterpiece, best quality, ultra detailed, etc.The higher the score, the higher the quality.Most people only use score_9, score_8_up, score_7_up it is enough to make the results better, but there is nothing wrong with using all these tags. if we include all score tags, it will make the results more diverse.There are more special tags such as"source_pony, source_furry, source_cartoon, source_anime"These tags are used after filling in the score tags, this aims to make the model focus on a certain style or thing, it is recommended to only choose one of them.Last is rating for descriptions or events with the character you create, mostly used to avoid explicit results/NSFW things."rating_safe" : this is used to avoid explicit things."rating_questionable" : this tag sometimes produces explicit results, sometimes also safe"rating_explicit" : used for explicit things and adult content.Some new models say no score. but if so, wouldn't it be called SDXL rather than Pony?As long as the model category is called Pony, I think the score tags is are still needed.This is a comparison image between using a score and not using it.ClosureOver the past few months I've been surfing in places where people post their art, I've noticed that some people just remix people's posts without understanding the prompt. This kind of thing will damage the ecosystem and result in not knowing how to write the right prompt.For example, someone creates art using a pony model. you know very well that pony models require score tags, but that person doesn't use score tags and he's lucky to get a good seed, then people remix the post and it spreads.If it's like that, it will have a bad impact on the owner of the model. indirectly, they represent the model with a image that shouldn't be like that.This guide is just a version of me. you don't have to follow everything written above. everyone has their own way, right?It's all based on my experience in the AI world for over a year.I will continue to learn new things. if there are any mistakes in the explanation in this article, you can correct it in the comment column. I would be very grateful for that.Lastly, if anyone says this article has nothing to do with Halloween2024, what would you expect from a Halloween guide article in here?You're not going to tell me to write an article about how to carve a pumpkin in here.
9