ApacheOne
/

Hunyuan_Part_2

V2 is the same dataset as before, just recaptioned with some revised training settings. It seems to offer a clearer picture, though not sure if it follows prompts any better. If you aren't getting the results you want, go back to V1. FEET VIDEOS!!! You may have never thought you wanted them, but now you realize you may just need them. I'm again not totally happy with this lora. It peaked at about Epoch 5 only to devolve into a mess for the entirety of the middle epochs, only resurfacing again to coherency at Epoch 15. Epoch 5 is usable, but when testing between the two, I decided that Epoch 15 was ultimately the better epoch. I will definitely try retraining this to see if I can recover a bit more stability. Trained to give you various foot focused poses. Use the trigger word as well as some of the below trained phrases. f33t The focus is on the woman's feet, which are positioned prominently in the foreground. A woman sitting on a floor with her legs extended forward, displaying her feet prominently. She flexes her toes and curls her feet, touching her feet together. The model is also trained a bit on the following clothing concepts: torn pantyhose. stockings. leg warmers. socks. cum on feet. Some of the sample videos utilize an unreleased lingerie lora I trained that may have helped with the above clothing. Not sure if I will get around to releasing that, so don't expect to match the look of the stockings/pantyhose exactly. this lora was focused on pose and motion. Use it at whatever strength you want. Trained on double blocks only, though at the 15 epoch level, it does tend to overwhelm character loras.

Civitai Links

File Information

Filename (Saved Locally): feet_epoch15.safetensors
Filename (Civitai): feet_epoch15.safetensors
Size: 172.55 MB
Hash (AutoV2): EF4445E60F

Downloaded locally using CivitaiLocalDownloader/1.4-SpecificVersions-NoCreatorFolder.