<ul><li>Attempting to solve the issue of excessive muscular legs when generating characters wearing leggings.</li><li>Training based on images of Asian influencers collected from social media, with no specific individuals or brand alignment.</li><li>Training based on chilloutMix, therefore, to get the best result, please use it.</li><li>Recommended weights for LoRA: 0.8 to 1, recommended CFG values: 3 to 6. LoRa</li><li>Generally, every sampling method and more than 5 steps can have a good result. Personally, I usually use 30 to 40 steps with the Euler a sampling method and DPM++ 2M SDE.</li><li>Use "(( [color name] leggings)" as the trigger word. Common colors included in LoRA can be found in the image below, but you can also use any other color, although the effect is not guaranteed. If you can't achieve the desired color, you can generate in gray and then inpainting with other colors.</li></ul><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/628d348e-ef26-4f5b-b433-279815286c0d/width=525/628d348e-ef26-4f5b-b433-279815286c0d.jpeg" /><ul><li>Choose different perspectives using trigger words “back view”, “side view”, and “front view”, “close-up” and “focus on [specific areas] ” to obtain detailed shots.</li></ul>

Files with SHA256: 744268b5e271a497ca711884c270aae4e533957e4861ba74729434e5849a57b1