Hello everyone, This is a requested LoRA of Suzaku from Heaven Death Game. Creating this model was quite challenging because I only had four images of her to work with. To expand the dataset, I used inpainting to generate additional variations and trained a Version 0.1 model first. I then used that model to help create this Version 1.0. The process is a bit tedious and involves some trade-offs. Version 0.1 is closer to the original character, but due to the limited dataset it tends to be less flexible. Version 1.0 offers better flexibility and can handle a wider range of prompts, although some details of her appearance may occasionally be slightly inaccurate. The model also has a somewhat AI-generated aesthetic, so using anime screencap LoRAs or other style-correction LoRAs is recommended if you want a look that is closer to the original anime/game.Note on the outfit: Due to limitations in the available reference material, I couldn't perfectly replicate the gold trim from the original game. Its design and placement may therefore vary somewhat between generations. <hr />Prompt: look at "Trigger Words" on the right side 👉Changelog: look at "About this Version" on the right side 👉<h1></h1>My Settings <a target="_blank" rel="ugc" href="https://github.com/Haoming02/sd-webui-forge-classic">(Forge <s>Classic </s>Neo</a>)*:*note, the example images only use high-res fix.Checkpoint: waiNSFWIllustrious_v<s>160 </s>170Sampler: Euler a, CFG Scale <s>7</s> 6Scheduler: <s>Automatic</s> KarrasSamplings Steps: 30Hires.fix: R-ERSGAN 4x+Anime6B, 20 Hires steps, Denoising Strength: <s>0.5</s> 0.45Extension:Adetailer (highly recommended)

Files with SHA256: 08243bec895610f311e1f8d65ec17508e671189b725ea3a5d38b207a874031b9