AI 생성 이미지 나눔 프롬포트 꿀팁 공유

Remember I said I was editing to the character's music!? It's done and I've uploaded it to KineMaster~! Come check it out through the link below!! https://x.kine.to/tT3amsl I'll post the prompt again in the comments!

I'll show you how the difference in training data I mentioned in my post appears in the results. It's roughly like this. Please understand that I can't share the images because I generated them for my 'peso-짤' (a type of meme/image) and intended to use them myself. I want to generate a woman holding a doll, but a random animal might appear. As you can see well in the photos, black cats, teddy bears, white rabbits, and gray cats come out well. These are tags with a lot of training data and good training for dolls. These are the ones that came out the way I wanted. Calico cats and white cats are quite ambiguous, right? Suddenly, the person hugging the cat becomes unrealistically chubby... This type of result comes from tags where the training for 'calico cat + doll' is scarce or non-existent. If you want to generate dolls, these are tags to avoid. For dolls, I think I used 'plushie,' 'fluffy doll,' and 'big doll,' and I remember adding '(desired animal) doll' or '(desired animal) plushie' in addition to those. I don't remember exactly because my records are gone.

You can use it as a doctor or researcher. I made this preset for a specific character chat, but since that character is on the verge of being deleted, I'm just releasing it. This also used tags like lab coat, white lab coat, lab jacket. Lab coats, research attire definitely work well with 'lab' tags. It's the same for backgrounds. Instead of just using 'hospital background,' using 'medical bay,' 'lab background,' or 'lab theme background' makes it feel more like a hospital or a lab. It's best to choose tags that are as specific and have a large amount of training data as possible.

I don't know what the problem is, but my post got rejected twice and I'm writing it again. Maybe it's because I added tags... I'll attach the remaining images and good tags in the comments.

Prompt Male, black hair, white eyes, wearing a sleeveless turtleneck and black cargo pants (with burgundy accents) as the base, and a vintage black baseball cap and a burgundy hoodie worn loosely over it. Pose him sitting on a rooftop, looking down. GPT even made a title for me lol Title: Rebel on the Rooftop Gazing Down at the Rainy City Our GPT is great at making titles too haha

Um.. I've become broke with pix ai.. I had over 1 million just a few days ago, but maybe I used it too much.. ㅠㅠ Or is it because I'm choosing candidates for hearts and coins..? Anyway, whether I save money, my images drastically decrease, or the quality drops, it seems like it could happen..

I think you can refer to it as 'doctor peso'. By the way, for doctor's coats, tags like 'Lab jacket' or 'white lab jacket' work better than ambiguous ones like 'doctor's jacket'. It seems to be a difference in training data.

Spreading Maid outfits I don't know if the LoRA or model I'm using was trained on 🔞 images, but it keeps exposing panties (...), so I had to put in negative prompts and rerun it countless times, what a hassle... I remember using various tags like 'girl', 'cute girl' in the front part, and for the latter part, I think I used the 'woman' tag with 'young girl' in the negative prompts. When generating images, it randomly exposes itself, and if a child-like figure is generated, there's a possibility of it being contaminated with loli/shota, etc. So, based on nai, please give priority to the 'aged up' tag or negative weighting for child-like figures. For Fix, you'll likely need to modify both negative and positive prompts.

This is the character I'm using for my Zeta user profile. It's a Grand Mage concept, but what kind of prompts should I use to make it look like a mage?! Even when I put 'mage' in, it doesn't work, so I'm curious how to create it. Please help me ㅜㅂㅜ

Prompt Woman, black hair, red eyes, half-up ponytail hairstyle, little devil-like style, small devil wings on the back, semi-realistic art style Suddenly got really into devils.. lol