There are 2 versions of this Lora. 2nd version will be released soon.Style Accurate isn't very flexible but it really captures the art style. Anime is more flexible but loses the original style.<ol><li>Style Accurate - trained exclusively on screenshots from episode 1 and 2 - using the AnyLora checkpoint</li><li>Anime - trained with screenshots and fanart - using the Animefull-final-pruned-fp16</li></ol>Full training details are in the "About this version" menu.<hr />Like what I do? Please consider supporting me ☕ <a target="_blank" rel="ugc" href="https://ko-fi.com/citronlegacy">https://ko-fi.com/citronlegacy</a>Any amount would be awesome!

Marcille Donato | マルシル・ドナトー (Dungeon Meshi)

Trained on: 71 imagesTraining Model: AnyLoRA_noVae_fp16-prunedflip_aug: FalseNum of Repeats: 10Unit is Epochs or Steps: EpochsNumber of Epochs or Steps: 10Training Batch Size: 2Total Steps: 3550Resolution: 512Network Dim: Dim=32; Alpha=16Lora Creation Process took: 0:37:25

Files with SHA256: df87f8844a1f9b83ed6b02d1bbb8d3ffb1bc8869170442fde4f5ed75cd520659