I made this LoRA to give generations a dynamic and epic look, reminiscent of the Genshin characters' splash screens! It works really well with both charachters, character LoRAs and landscapes, if you want to spice up some fantasy setting. VERSION 2 I retrained the LoRA with upscaled images and revised the trigger words since "splash" gave it a strong bias towards water. Now the triggerword is "gs2", but the LoRA also works well without the trigger. The LoRA is now more flexible and easily steered towards the desired result. Faces are still wonky, but you can fix them with inpainting (or automate the process in Automatic1111 with the extension "after detailer") You can still use words like "swirling" "dynamic" or similar to give it more dynamism. ------- VERSION 1 Use its trigger words and a description of the background splash theme to bias it towards different colors or elements, for example you can append to your prompt: "red_theme", "purple_theme", "magma, fire tornado, red_theme", "ice crystals, blue_theme" etc. and play with the weights of your tags if necessary (refer to the example images). It is possible that the faces are not so well generated. This is normal since this LoRA focuses on the background. To improve the faces, just inpaint them :D This LoRA also likes water quite a bit, so if you need, put something like "(water, blue, waves:1.5)" in your negative prompt ------- The image examples i posted contain: - simple prompts with character description + color/element theme - this LoRA mixed with other character/style LoRAs i trained The samples are generated with the following negative embeddings i found here and there on the internet: bad-artist, bad-hands-5, bad-image-v2-39000, bad_prompt_version2. If you like this model, consider offering me a coffee! :D <a target="_blank" rel="ugc" href="https://ko-fi.com/erentil">https://ko-fi.com/erentil</a>

GenshinSplashCharacter-Style

I retrained the LoRA with upscaled images and revised the trigger words since "splash" gave it a strong bias towards water. Now the triggerword is "gs2", but the LoRA also works well without the trigger. The LoRA is now more flexible and easily steered towards the desired result. Faces are still wonky, but you can fix them with inpainting (or automate the process in Automatic1111 with the extension "after detailer") You can still use words like "swirling" "dynamic" or similar to give it more dynamism.

Files with SHA256: 6679971746c3a32866d3af0dc6883fbdf50441e42f5f3e2cab7a1181d077b54c