This LoRA is based on official art for the Magical Girl Lyrical Nanoha franchise. I have focused exclusively on the young protagonist, Nanoha, and carefully constructed a training dataset consisting of high quality art and tagging. Over the span of several weeks, I trained and retrained to accurately diffuse a style similar to the official art style. The diffused images should be immediately recognizable for any other collectors of the official Nanoha artworks!

Tagging

The activation tag is 'takamachi nanoha'. If you want her most recognized appearance, add 'twintails' and 'white ribbon'. This is the way she is most often depicted. But the training dataset consists of other hair ornaments and hair styles, so feel free to experiment here. For example, try 'straight hair' for her nighttime hair or 'black ribbon' for Fate's ribbons!

Strength

The strength of the LoRA should be anywhere between 0.6 and 0.8. If you prefer the visual style to match the official art as much as possible, stick with 0.8. The trade-off for this is losing some prompting flexibility. Lower the LoRA strength to regain that flexibility! Anything below 0.6 loses her appearance and the official art style.

Quality

Preview images here on civitai are heavily compressed and introduce quite a few JPG compression artifacts. To see the true quality this LoRA can accomplish, generate one of the preview examples locally and upscale with R-ESRGAN 4x+ Anime6B. Or, click here: https://i.imgur.com/SBQoruy.png.

Feedback

I would love to hear your thoughts and see your diffusions! Please share them~