Type | |
Stats | 156 4,628 1.5k |
Reviews | (24) |
Published | Sep 17, 2024 |
Base Model | |
Training | Epochs: 8 |
Usage Tips | Clip Skip: 2 Strength: 1 |
Hash | AutoV2 7E42614CBD |
V5aLTER:
V5 alternate with different training params and modified dataset tags, attempt to remove some of the 3d influence from POV images
Different enough in style that I'm keeping V5 around in case people prefer the original
Seems to have slightly better consistency and anatomy and small details
V5:
V5 trades some of the extras and nice-to-haves of 4.21 for consistency, versatility, and environmental variety: Not as good for comic gen, better for prompt adherence, texture, eyes, hands, and clarity!
To see just how versatile V5 is, ALL images shown in slideshow above were created with Pony V6/Anime Model OtakuReliableEnable + my V5 Lora exclusively.
I recommend CFG 6-9, Euler a/DPM2 a, Strength 0.7-1.0
Goes INCREDIBLY WELL with Anime Model OtakuReliableEnable (AMORE) base model, find it in Suggested Resources below
Overview:
Good all-around Lora made from a lot of different styles and types (3d, anime, realism, cartoon).
GREAT FOR SHINY SKIN/OUTFITS AND OUTFIT DETAIL
Can do bold lines well, in addition to soft edges/gradients like those seen in 2.5d/3d images. Almost like a Reiq-style for 3d, Jojo's for cartoon/anime, flat for anime, but also flexible.
Works well with most Pony types.
Seems to work well and mix with other Style Loras without issue.
Fishnets, bodysuits, and skin tight clothing come recommended.