The V2 version significantly reduced the number of outdated images in its training, resulting in a style that is more stable.
Although it's not necessary, but adding "clean soft color" while prompting seems to have better results.
