Takes only 1 minute. Beginners will miss this trick. It will vastly improve the quality of all of your art generation. Note: this only works for most anime-based models. Realistic models will not be affected.
This was done using the AbyssOrangeMix2_sfw stable diffusion model without a VAE.
Technical details:
Steps: 34, Sampler: DDIM, CFG scale: 8, Seed: 2007333899, Size: 512x512
Prompts are:
girl, blue dress, portrait, masterpiece, twintails, street
Negative prompts:
(low quality, worst quality:1.4)
Please like and subscribe to receive the content that I have planned on my roadmap. It'll be an interesting journey!
![](https://s2.save4k.ru/pic/2fzTVogW2Zc/maxresdefault.jpg)