Anime Screencap Style
NEW ERA v1.0 (SDXL / PONY DIFFUSION version combining almost all my popular models with an emphasis on retro anime)
P.P.S. new model Anime Screencap / LORA / PONY DIFFUSION and NEW ERA 3.0 (MODEL) PONY DIFFUSION / SDXL on Boosty
I made a video on how to achieve the same quality or just repeat my art
P.S. version 1.5 (2.5 times more dataset, more NSFW, more details and better quality) on Boosty
P.P.S. new model ogipote (2.5D)_v3.0 and ogipote (2.5D)_v2.0 on Boosty
version 1.4 is an attempt at high widescreen resolution, there should be less distortion, but this is not a panacea
From the name it is clear that the model is designed to implement screenshots extracted from the anime. Why a model and not Lora? I still think that the model is more versatile and gives much more "correct" results, and frankly the model is capable of better quality than Lora. However, I do not deny that I will make Lora according to this model, if there is interest.
VAE is embedded
negative promts: (worst quality, low quality, extra digits:1.4) or (worst quality, low quality:1.4)
Eta noise seed delta 31337 (if you want to repeat the art)
Trigger Words (prompts):
anime screencap
anime coloring (optional)
almost all the art is made on a hires fix with the following settings: upscale by 2, Denoising strength 0.57 - 0.6, upscaler latent (neares-exact), the original resolution is 512x768 or 576x832, or 640x960
CFG Scale - 5 (it's very important guys, I noticed many people use 7 by default, they don't need it, since vae is built in, then on 5 the image turns out juicy, no more is needed. In addition, the hands and proportions are obtained better by 5 and the characters do not look like "overcooked chickens")
Please post your work with or without comments, it will help me improve. Thanks!
If you like my work, click on the heart above, I will be pleased :3
Description
enlarged dataset
attempt at high resolution (widescreen images should look better)
more accuracy
FAQ
Comments (8)
didn't you use fp16? the model weighs a lot
bp16 train and merge with my others models.
Of course, this is not an ideal model, but it is definitely very good! It is quite flexible and diverse, and overall stylistically very cute. I like it!
Everytime I use the word "anime (screencap", it creates child-like faces. If I use a negative embed "child-face", it just adds more. Is there anyway to get the screencap effect but not the child-like face for characters?
it's weird, I didn't have such problems, but if they haunt you, use booru tags. Negative: loli, child
In regular prompts you can write breast size, for example: large breasts, medium breasts. This is just an example of solving the problem.
All my models and LORA (without exception) are built on booru tags, if you study them, it will be easier for you.
Thanks! will do
I restored my PATREON (or rather created a new one with the permission of patreon), if someone wants to support I posted there all my new models and LORAs, which were paid on BOOSTY in one archive. I will be glad if you even subscribe to patreon for free, I will know that you care about my work and you want to see updates.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.