Hunyuan:
Bringing a bit of diversity to the other loras. It's picked up the style pretty well, but you can boost it by adding attributes typical to the time and genre.
Wan:
Definitely not the easiest model to work with. I retrained it 3 times and increased the dataset. It still struggles to get rid of the "wan look". But it can produce some nice results, too. Just requires a bit of work. Check the samples for prompts and settings.
Description
FAQ
Comments (10)
I hope you can share your training experiences. It sure seems like you're learning some good lessons in training and retraining, and the results of this one are the best yet.
I'd love to learn what went wrong, what went right, what kinds of training captions you wrote, how many videos, settings, etc... but I know that's a lot to ask for.
Yeah, I'll try to put together a guide. Just need to find the time :)
my man still the only one on this porn site actually doing good style loras for my favorite model. bless you
It is a porn site... that's actually sad.
@null i love that everyone is free to do whatever they want--and the lack of censorship is why i love it here-- but it gets tiring to have state of the art image and video tech and its nothing but porn, i mean at least give me a 50/50 split lmao. the ratio of porn to sfw is ridiculous and overwhelming sometimes
I can hear it. I can hear the "DdDdDdDAAAMN!" and sassy one liners.
Can this do Fly Guy with the fish shoes?
The Fine Trainers Hunyuan one looks better to me than the Wan 2.2. If there is a way to make it work with Comfy and retain the quality that would be good.
I agree that the hunyuan lora's usually has better style adherence. But it might be due the training regime (check the difference for the recent 80s fantasy lora). I can upload a comfy compatible version if you'd like. No retraining needed, just conversion.
@neph1 Okay. Thank you.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.