[Offical Game Style] Genshin Impact Offical art+Profile picture

· Based on what I consider the optimal/most representative facial content for Genshin Impact's overall aesthetic.

· I've removed most elements that could cause LORA description contamination (excessively character-specific pupil shapes and gradient irises, hairstyles, and hand-drawn-style content from early art directors), so the style may lean more towards the effects seen after 3.0~4.0.

· Most male character images are too blurry, and their facial shapes may contaminate most female characters (I don't believe LORA distinguishes between youths and young adults). Therefore, I may create a dedicated Genshin male character LORA in the future.

· The source material primarily comes from large-scale illustrations released at the game's new updates and some character portraits. If you have any recommended source content, feel free to leave a comment under this LORA.

· TIPS: Testing has shown that it is only suitable for foundational models trained toward an animated/cel-shaded style. For unknown reasons, content generated by haruka and hoshino may not accurately reproduce LORA's style and may even forcefully alter it.This issue has been resolved in version 1.5. You are free to use the haruka2.0 model to generate more content. For version 1.1, please continue to follow this advisory.

Recommend tags combo: Genshin Style,anime style, (cel shading:1.2),anime coloring, high saturation, clean lineart, soft ambient lighting, glossy highlights, smooth gradients, vibrant lighting, detailed rendering, offical art,bright color palette, 8K wallpaper,masterpiece, rim light,outer glow (1.5 This prompt can be ignored.)

· Updated to 1.5: It has been confirmed that no watermarks exist in the original training dataset. If low-quality images such as watermarks appear, this is not an issue with this LORA. It is likely caused by the base model being trained on images containing watermarks.

· If you wish to fuse other LORAs during use, please utilize version 1.1 for optimal results. Due to the greater stability of training materials in version 1.5, incorporating additional LORAs may lead to subpar quality or over-reinforced content.