<h2 id="v5alter:-q5ju3bfb0">V5aLTER:</h2><ul><li>V5 alternate with different training params and modified dataset tags, attempt to remove some of the 3d influence from POV images</li><li>Different enough in style that I'm keeping V5 around in case people prefer the original</li><li>Seems to have slightly better consistency and anatomy and small details</li></ul><h2 id="v5:-qhvfivbcz">V5:</h2><ul><li>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!</li><li>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.</li></ul>I recommend CFG 6-9, Euler a/DPM2 a, Strength 0.7-1.0Goes INCREDIBLY WELL with Anime Model OtakuReliableEnable (AMORE) base model, find it in Suggested Resources belowOverview:<ul><li>Good all-around Lora made from a lot of different styles and types (3d, anime, realism, cartoon).</li><li>GREAT FOR SHINY SKIN/OUTFITS AND OUTFIT DETAIL</li><li>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.</li><li>Works well with most Pony types.</li><li>Seems to work well and mix with other Style Loras without issue.</li></ul>Fishnets, bodysuits, and skin tight clothing come recommended.

CicaStyle [PonyXL]

Quick fixes for eye blurriness and hand/anatomy issues. As a bonus, 4.2 can do "comics" with some actually decent sequential flow, due to added manga/comic scans in training. Scaling and weight are a lot more precise, with weight 1.0 giving best results at a CFG from 6 and up to 12!

Files with SHA256: 4a6e96ac7c45e910a3599796381ccb3abbbbdffc4b3a8b693ff8613ce503f731