<h1>Fervent Tempo ∆</h1>Character: Rozaliya Olenyeva (Honkai Impact 3)Battlesuit: Fervent Tempo ∆, original outfitPixiv post <a rel="ugc" href="https://www.pixiv.net/en/artworks/107066054">here</a><h1>Usage (Important!)</h1><h3>Compatibility</h3>Should work with most anime-like checkpoints. Tested with <a target="_blank" rel="ugc" href="https://huggingface.co/WarriorMama777/OrangeMixs/blob/main/Models/AbyssOrangeMix3/AOM3_orangemixs.safetensors">AOM3</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/WarriorMama777/OrangeMixs/blob/main/Models/BloodOrangeMix/BloodOrangeMix.safetensors">BOM</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/models/2583">Hassaku</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/models/24383">Grapefruit</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/models/15509/aoaoko-pvc-style-model">PVC-AOAOKO</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/models/5415/cornflower-x-stylized-anime-model-wnsfw-support">Cornflower X</a>, etc.<h3>Weight</h3>Recommended 0.825; 0.75~0.90 range works nicely depending on how much flexibility you need.<h3>Trigger Words</h3>No trigger words are required. However:You can put (mismatched stockings, long hair:1.15) on prompt and (tail, mechanical tail, weapon, sword, holding, holding weapon, holding sword:1.15) on negative prompt for better consistency.<h1>Other</h1><h3>Image Details</h3>Preview images use AOM3 or BOM, unless stated otherwise.VAE used is <a target="_blank" rel="ugc" href="https://huggingface.co/WarriorMama777/OrangeMixs/blob/main/VAEs/orangemix.vae.pt">orangemix </a>(nai)<h3>Training Details</h3>May vary based on version. See individual versions for details.

Fervent Tempo Delta ∆ (Rozaliya Olenyeva) (Honkai Impact 3)

Probably not the best, I'm still learning and experimenting with LoRAs. I just thought this seemed "good enough" to upload somewhere.Outfit changing is probably not possible easily due to limited training data. (they're all trained on her original attire on this version.)I forgot to put _Release1 in the file name, I think. Oops. Subsequent versions (if any) shouldn't have this issue.Training Details (w/ bmaltais/kohya_ss)Base checkpoint trained on: animefull-final-pruned (NAI)14 images (See Training Data file), 30 repeats, 2 epochs (840 steps?)Learning Rate: 0.00032LR Scheduler: Cosine w/ RestartsText Encoder LR: 0.00008U-Net LR: 0.00032Network Dim: 128Alpha: 64

Files with SHA256: 0771a829c21697451963921cb937e14b99cc49673e91517ec6d8d3e880f5c0f0