This is a v-prediction model, please make sure you're using a compatible UI.
I wanted to make a merge that gave a solid 2D anime screenshot base style without needing to throw in a ton of extra tags. See galleries for my recommended prompt structure, all of my images are a single pass with no upscalers or refiners.
All credit goes to the original checkpoint authors and whoever they credit for their own models.
Recommended Settings
V4
Steps: 20-30
CFG: 3-5
Sampler: Euler Ancestral
V3
Steps: 20-30
CFG: 2-3
Sampler: Euler Ancestral
Description
Updated the mix to hopefully get more varied poses and better lighting. Let me know if you think it's an improvement or not! Check the gallery for prompt structure, keep negatives simple for the best results.
FAQ
Comments (26)
Thanks so much I can start my kyōzon anime now 🫂😍😊
Oh man awesome!!! can we expect an update on the non-vpred too? I seriously love this model sm!!!
Hi! I'm experimenting with it now, if I think it turns out well I'll publish an update.
@aimalias you are the best! thank you so much
I've reviewed 3.2 here (Japanese, but Translate will get through fine):
https://www.hoshikou-ailabo.net/blog/2025/06/15/anime-screenshot-merge-review/
Thank you for taking the time to review my checkpoint! Sorry to put out an update as soon as your review was done!
Extremely grateful for every great update you share with us, Aimalias!
Thank you very much for everything!
Cool checkpoint!
Some initial thoughts comparing it to 3.1 and your Illustrious 3.0 version (Euler A / AYS / 30 Steps / CFG 3.5, "anime screenshot, anime screencap, anime coloring" prefix):
- Your 3.0 Illustrious and 4.0 Noob models are easily the best "anime screenshot" models I've seen so far.
- Illustrious 3.0 version is probably the most "accurate," while Noob 4.0 is more "idealized" screenshot. The Illustrious version produces images that are slightly blurred and soft, as if viewed on an older TV. The Noob version is crisper and clearer, but still "feels like" looking at an anime screenshot (think HD remaster/cleanup version).
- The Illustrious 3.0 version is slightly worse at depicting nonhumans.
- Noob 4.0 vs Noob 3.1 is better at maintaining a consistent look and feel across different prompts. Lighting also feels significantly improved.
- Prompt adherence seems comparable.
- Newer models are more prone to adding captions/subtitles than older one.
Thank you for comparing everything!
good model! gooooooooooooooooooooooooooooood!!
I think with v4, you do get more poses, but they don't always look very dynamic. Nonetheless, there's no doubt it is better than v32 in that aspect, where sometimes it felt like a loop with more complicated prompts.
I will keep using v32 only because it retains the style elements of certain characters better.
hi! I haven't had any issues with style consistency, but I include the character + series name tags, official art also helps
@aimalias where do you get the series name tags like "nami \(one piece\)"
@gaara675 Danbooru!
@aimalias do you get the tags from random artwork or official
@gaara675 I just add "official art" as a tag to get it closer. Can also use the OG artist like eiichirou oda.
Could anyone, please, give any advice how to avoid of appearance of other's person body parts?
i.e. if in promt there is "grabbing own butt" - high chance that it will be grabbed by other person hands. I've tried to use "crowd", "other characters" in negative promt, but it seems not to work every time
Include "solo" in the positive if you didn't already try, right after the character name!
solo in pos, pov in neg
1girl,solo,grabbing own butt by self
Underrated checkpoint
Honestly I wasn't going to comment but This model is one of the Few models that actually does what it intends to do 99% of the time. It makes images really close to the source material
Really good model. Sad that I didn't find it eariler and needed to use WAI-slop and cyberfix (with blinding bright outputs).
Very good for my character lora because characters usually from anime and hentai.
I don't need it. I Don't need it. I definitely don't need it. I don't need it. I NEED IIIIIITTTTTT!
The best checkpoint!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.