My seccond attempt at Video training, using Animations style of Mizumizuni, this was quite difficult to do on my limited vram. Im going to need a 5090 if I want to keep supporting things like this and make them properly.
Heres the Prompt ive been using: Mizuni, Pov of a girl performing oral on a man, partway through she stops as the man ejaculates in her mouth
you might need to make some changes to the prompt, its going to be listed as a v0.1 as it works but its inconsistent and its a lottery roll if you get something decent out of the generation.
future versions will remove the censor bars and make it more consistent.
This release is part of my cadence improvement strategy. I’m uploading functional LoRAs that don’t yet meet my full standards but still perform well. These versions are will be marked as experimental. Expect a steady flow of updates on my account, with continuous refinements to reach final release quality.
Description
FAQ
Comments (8)
how to use this lora?
you have to download and use Hunyuan video locally this lora is for that model
Great work man :)
Are you running hunyuan using comfyui?
Yes
@Midori2667 Could you post your workflow?
@crazymanjj id love to but I dont use Comf often enough so I keep redeploying it when I need it. which means I dont have it on hand,
If you run across an error with transformers library like "cannot import name 'xxx' from 'transformers.xxx'
You're most likely using a newer version of the library. The most up to date version of the transformers lib you can use is version 4.47.1.
For those of you not using a virtual environment you can simply run.
pip install transformers==4.47.1 --force-reinstall
if you are using a virtual environment then entering that environment and replacing the version is probably the best option.
You could also most likely update this in the requirements.txt but the solution above worked for me.