<h2 id="updates:-cd7mmkdfe">Updates:</h2>🔹 Uploaded V2 and sample gallery<h2 id="details:-y40gsxevg">Details:</h2>🔹 Trigger Word(s): evulchibiart style🔹 Trained On: <a target="_blank" rel="ugc" href="https://civarchive.com/models/257749?modelVersionId=697402">PonyDiffusionXL v6</a> (Base)🔹 Recommended Strength: 0.8-1.0🔹 Training Samples Seen: 8000This is a (IMO) very high quality trained lora on around <s>270</s> 290 pieces of HD artwork, with <s>three</s> eight runs to find good hyperparameters. It's extremely good at pinup and portraiture of course, but will struggle to have non-humanoids in the images unless you try very hard in the negative prompt.<h2 id="versions:-su1few2cj">Versions:</h2>🔹 v1: Self explanatory. Baseline finetuned lora, works best with PonyDiffusionXL-v6.🔹 v1-NoTE: No Text Encoder version of v1. More flexible for use on SnowPony and other finetunes, style transfers more agnostically.🔹 v2: Retrained on data with cleaner results, no watermarks on images by default, more epochs.

Artstyle 🎨🖌 - EvulChibi [PonyXL+Others]

No Text Encoder (CLIP)A version of the lora with no Text Encoder (CLIP) weights, this is more flexible for non-base checkpoints and finetunes of Pony. This will be more flexible than the normal LORA.

Files with SHA256: a590f64127705d58f0f625f111498ed5dcb708e198ac0d68fd688cb7de9b7271