The Lora for converting anime to realistic human figures has been updated to Qwen image Edit version 2511. The planned steps for Lora are 30,000, and the current version is at 15,000 steps. The dataset consists of 103 groups of images, totaling 206 pictures. The final version will be updated in a few days!
Optimization list:
1. Optimized the consistency of facial features and the integration of prompt words.
2. Improved the restoration of facial expressions.
3. Enhanced the restoration of different demographics, such as children, boys, and girls.
4. Increased realism.
5. Supplemented the training set with realistic background restoration, maintaining outstanding performance in recreating authentic scenes.


2509 official version (base model) can be viewed here: https://www.liblib.art/modelinfo/1cf56301d60d440184bb907c56cfc75f?from=personal_page&versionUuid=a4d909d0ff024f20ad813a2fc1d6314f
Recommended parameters:
CFG:4
Steps: 30


Bilibili: AIGC特异点

Creator benefits: Click the avatar in the upper right corner, select "Invitation Code," enter the code, and receive 1,000 RH coins. Log in daily to earn 100 coins.
Invitation code: sdhs0trb

Bilibili: https://space.bilibili.com/49766729?spm_id_from=333.1007.0.0

Youtube tutorial: https://www.youtube.com/@virtualAIcoser

For technical discussions, leave a message on Bilibili or in QQ group 821194069.