Type | |
Stats | 952 |
Reviews | (142) |
Published | Sep 12, 2023 |
Base Model | |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 DAF23BA3D9 |
[AingRealistic is no longer relevant and no longer maintained. I lost interest in realistic models. Thank you for using, tho.]
===
AingRealistic is a merge between AingDiffusion and some of realistic models. The goal of this model is to generate something between anime-styled images and realistic images.
Guide to generate good images with this model
The default negative prompts I use are: (worst quality, low quality:1.4), monochrome, zombie, (interlocked fingers:1.2). You can use Textual Inversion like EasyNegative too, but I personally rarely use it.
Recommended samplers are "Euler a, Euler, and DPM++ SDE Karras". You can try other samplers but mostly they didn't work as good as these.
Hi-res fix is a must if you want to generate high-quality and high-resolution images (and eye color fix too). For the upscaler, I highly recommend SwinIR 4x or 4x-AnimeSharp.
Set Clip skip to 2, eta (noise multiplier) for ancestral samplers to 0.667, and ENSD (eta noise seed delta) to 31337 (ENSD is optional, but needed to replicate my results).