The results were a bit worse than I thoughtLater I'll try to retrain on CogVideo 1.5 or the new 2B neural network LTXVideo (If it has the ability to train LoRa)The prompts were automatically generated by CogVLM so they may not contain nsfw notations.Here is an example of the prompt used.shirt_lift. slow motion. The animated 3D female characters are dressed in lingerie, initially appearing wearing a bra. They then slowly lift their bras, maintaining the same outfit and setting. By the end, they are standing with their bras lifted, proudly displaying their bare breasts, surrounded by a background that highlights their animated features and the bright colors of their outfits.If anyone comes up with a more stable prompt with decent results, please write it in the comments.

Files with SHA256: f8fb1e1a39673fc9824981c266a2e6300daccbca5d869bf81c4c177cc8a7b967