Recommended weight: 1.0-0.6 Typical image generation models may not interpret “downturned eyes” correctly, resulting in expressions such as “upturned eyes.” This LoRA supports the generation of correct downturned eyes.