This is my first Minimax-H3 lora model, which I trained with a lot of effort and achieved excellent performance in FL2VA.
I will update 30 NSFW models on Civitai.
Completed/Incomplete: 1/30The models used are as follows:
[diffusion_model]: DasiwaMinimaxH3_dasiwaHybridV1.safetensors
[lora]: minimax_h3_turbo_4_DasiwaREF2VAHybridV1_curveproj1025_compat_v001-T8.safetensors
[lora]: H-KJ.safetensors
Sampling
steps: 4
sampler: dual_clock_euler
scheduler: native_flow
If you have any good ideas, please let me know in the comments section.
If you can, please give me a like; it will give me more confidence to train more models.
Description
The first carefully trained model
FAQ
Comments (4)
Good job. Did you distill the oral insertion wan lora?
There are many ready-made LoRa modules for WAN, and currently I am mainly focusing on H3
@smNime My question was if you used the wan oral insertion lora to generate the dataset for the H3 lora training. The concept is very similar. Because i did the exact same thing with a dataset of 70 videos but never had the time go get into H3 training.
@FrogOnStilts Because it doesn't matter for H3 as long as the movements are fine.