-----------UPDATE-----------
I've retrained v1.1 with the I2V model. Same dataset but the movement seems to be much improved. Use this version for I2V, stick with v1.0 if you use T2V. I've generated with some of the same images as v1.0 so you can see the difference.
-----------DETAILS-----------
This lora was trained on 25 videos for 184 epochs. The prompts I used are added to my samples, but it's fairly simple, such as:
A ???? woman lying on a bed. She inserts two fingers into her ?????. She ??????????? by sliding her fingers in and out of her ?????. It should also work from a kneeling/all fours position but I've not tested this yet.
It also pairs quite well with my ???????? helper lora (see suggested resources below), at a strength of 0.4-0.6. Help the ?????? not get distorted so much as the fingers move in and out.
Trained with the I2V model for improved movement.
1. 转载模型仅供学习与交流分享，其版权及最终解释权归原作者。
2. 模型原作者如需认领模型，请通过官方渠道联系海艺AI工作人员进行认证。我们致力于保护每一位创作者的权益。 点击去认领
