I was asked recently how to get higher quality images like the examples and about blurred eyes and faces:īlurred eyes, or sometimes warped faces are not that uncommon in SD and sometimes require an inpaint to pull off. Here are some example image comparisons with some of the most common / top anime oriented models and also my previous model Dungeons N Waifu’s v2.2: Play with it and see what works best for you. Notice that the art & artist style of the prompt comes out more in 1 and 2 has more of the industrial / cyber aspects. (Unsure if this is directly affected by LORA’s will have to test more) Example: However… it will also work with Clip Skip 1 better for prompts that are going after a specific art or artists styles. (Look I didn’t plan this… the merges just made it that way.) I suggest a (((kids))), (((child))), (Young:1.5-2.0) in your negative and a (mature adult:1.5-2.0) in your positive. It does styles and illustration, and artists based prompts pretty damned well and adapts to most LORAS with ease, it DEFINITELY leans a bit towards nudity / NSFW and also more towards younger looking girls. Backgrounds that look like high quality anime or 3d rendered, game backgrounds. Specializes in illustrated, semi-realistic, painterly, subjective reality and a 2.5d aesthetic with semi real to real backgrounds. It doesn’t do everything but what it can do off the bat is a pretty damned amazing images in anime styles 2 - 2.5D quite effectively. _ (Older Info from version 1.5) _Ī handpicked and curated merge of the best of the best in waifu’s, anime, 2D, and of course, fantasy environments, characters & creature designs. Upscaler Remacri Upscaler, Lollypop Upscaler, 4x_NMKD_DigArt_10k| Hires Steps Equal to Half or Exact amount of Steps or higher Sampler DPM++ 2M Karras, DPM++ SDE Karras, Euler A, DDIM, Here are some optimal example settings for you to start with, but please keep in mind YMMV! This is a complex merge so even the pruned is also large as well. Note: There is a full & pruned version of this model available.
*** THESE SHOULD BE EASILY REPRODUCIBLE RESULTS! IF YOU HAVE QUESTIONS PLEASE ASK □ In most cases untouched other than mild inpainting for the eyes or face in a few examples. *** Note: that the example images are provided as a basis of reference and raw output. (I typically use (bad-hands-5) & (bandhandv4_anime) at the same time to cover more bases)
Always good in a bind)īadhandv4_anime (absolutely your go to neg for anime hands)īad-hands-5 (good for the more realistic images but will not work amazingly for anime)īad-artist \ bad-artist-anime (I’ve had good and bad luck)īad_prompt \ bad_prompt_version2 (works very good sometimes)
I recommend using the following negatives:ĮasyNegative (The all around catch all. There is no embedded VAE but it was designed to be used with this: This is a complex merge so even the pruned version is large as well. There is a full & pruned version of this model available. READ ME! THERE IS INFO ON HOW YOU CAN GET THE MOST OUT OF THIS MODEL! ***USE CLIP SKIP 2 FOR THIS MODEL! YOU CAN DO CLIP SKIP 1 AS WELL BUT 2 IS BEST!*** Very variable and very good at fantasy! w/ amazing backgrounds! It’s still a bit flavored to younger faces so (mature adult) may be necessary.Ī bit less NSFW heavy but you can keep it in the negative to be sure.