<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)

Addresses some attire issues from Release 1~3!!Improvements (vs. Release 3):Stockings now appear blue and pink much more reliablyThe right side of her jacket (from viewer's perspective) is now black much more reliablyTrigger WordsPut (long hair, mismatched stockings:1.15) in positive prompt, and (weapon, sword, holding, holding weapon, holding sword, tail, mechanical tail:1.5) on negative promptThis is not required but seems to yield better results.Training Details (compared to Release 3)The training images are the same, but one image was flipped horizontally, and another was moved to the lower weight folder. Also weight for the higher weight folder increased a little.Training Parameters(w/ bmaltais/kohya_ss; see Training Data File for full LoRA training parameters extracted with kohya-ss/sd-webui-additional-networks)Trained on: nai.ckpt (925997e9)Training Batch Size: 2LR: 0.0007, UNet LR: 0.0007, TE LR: 0.000175, Scheduler: Cosine w/ RestartsEpochs: 3, Dim: 32, Alpha: 16, Resolution: 512x512, Clip Skip: 2*EasyNegative in the example images is an embedding from <a target="_blank" rel="ugc" href="https://huggingface.co/datasets/gsdf/EasyNegative">here</a>.

Files with SHA256: 870ba1929f1bbdf2193d854798aba3f6571c9fa6bdf4779713d2da25e25243d5