Cute RichStyle 1.5

Cute RichStyle - 512x512 Model trained in SD 1.5 with photos generated with Midjourney, created to generate people, animals/creatures...

You can also make objects... landscapes, etc, but maybe you need more tries:

30 steps - 7cfg

euler a,ddim, dpm++sde...

you can use different resolutions, you can generate interesting things

TOKEN: cbzbb, cbzbb style, cbzbb style of _____ , you can put the token , (it is not required) but it is better to put it. Many times the token between () works better

possible positives: cute, little, baby, beautiful, fantasy art, devian art, trending artstation, digital art, detailed, realistic, humanoide, character, tiny, film still of "" , cinematic shot, "" environment, beautiful landspace of _____, cinematic portrait of _, cute character as a ""....

most important negatives (not mandatory but they help a lot) : pencil draw, bad photo, bad draw

other possible negatives: cartoon, woman, man, person, people, character, super hero, iron man, baby, anime...

((When you generate the photo, there are times when it tries to create a person/character, that's why the negative character prompts etc...))

landscape prompts better between ( ) or more parentheses, although it is not always necessary

you can use other styles, removing the "cbzbb" token and adding pencil draw, lego style.. watercolor etc etc, it doesn't make the exact photo style with which I trained it but they look great too!!

Most of the photos are daytime, to create nights it once worked with:

positive: (dark), (black sky) (dark sky) etc etc

negative: (blue day), (day light), (day) (sun) etc etc

Improve images and increase resolution Generate images and choose the one you like the most, and send it to the img2img section.

In img2img, choose to generate 2-4 or more images. If you want it to resemble the original photo, set "denoising" to 0.4-0.5-0.6. If you want more changes in the image, you can set "denoising" from 0.6 to 1.

Once those 4 images are generated, choose the one you like the most (or generate more images until you find the one you like) and send it back to img2img.

Now we will increase the resolution of the image. If the photo is 512x512, we can set it to 680x680 and 50 steps for example (changing "denoising" if we want it to resemble more or less) and generate 2-4 more images again and choose the one we like and send it back to img2img and keep increasing the resolution of the image little by little until you want to, in order to not lose details.

At this point, we can lower denoising to 0.15 if we don't want the image to change, and generate 1 image with 50/100 steps, this way we can make giant images with high resolution/quality.

Change person/character in the image: if you like the photo but want to change the character, send a photo to img2img, change the name of the character or person or animal and between 0.7-1 denoising