This is my first style that I decided to publish. Its main feature is the emphasis on anime-style eyes and enhanced elements of realism.<hr /><h3 id="v1.0">V1.0</h3>Test version for the Zimage Turbo model. The style works perfectly without a trigger and produces excellent results that combine anime and semi-realism. The trigger enhances the main feature of the style — the characters' large and expressive eyes. The style focuses exclusively on female characters. This version is based on the original dataset. Standard danbooru tags were used as captions, which work perfectly with the style.<hr />

Shiro Aurora Mix Style (ZIT)

LoRA works without a trigger, but the trigger enhances the main element of style — expressive and larger eyes.

Shiro Aurora Mix ZIT.safetensors

v1.0 ZIT

This is my first style that I decided to publish. Its main feature is the emphasis on anime-style eyes and enhanced elements of realism.<hr /><h3 id="v1.0">V1.0 </h3>Should work best with anime checkpoints, hybrid checkpoints, or checkpoints aimed at semi-realism. My checkpoint ShiroMix (v1.5) or checkpoint the WAI-NSFW-illustrious-SDXL will work fine. When using the tag 'realistic' in my checkpoint, the model is tuned to a more realistic style, but the LorA effect is still noticeable.In some cases, the style will work better if you increase the weight of the model from 1 and above (for example in the hassakuXL model)This style is included in the second version of my checkpoint ShiroMix. When using LoRA with my second version checkpoint, the style effect is noticeable, but a stronger impact can be achieved by increasing the LoRA weights.

Files with SHA256: 04a9e24a41a99a9c3ef26cd1dc5acfb24d9ab0ea273a2a663f53bf789dcf6b31