>>72300499So this is my first attempt of training on Ponymagine and running inference on autismconfetti. From the looks of it, the clip in the Shiori version of ponymagine is not enough to translate information that well. The lora requires 2 weight for unet and 5 weight for TE to get something looking like Ponto. However the result is close. I would say I need to try making a training merge of Ponymagine for training with higher TE mixed in, for the resulting loras to work properly. Haven't tested if it works on AniXL yet but it should work fine.
