Instructions to use MiniMaxAI/MiniMax-H3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use MiniMaxAI/MiniMax-H3 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("MiniMaxAI/MiniMax-H3", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
my opinion about nsfw
i tested in my gpu A100 80.model looking too good.
good face consistency.
less censorship.
understand the body structure perfectly.
prompt following is too good.
skin texture,slaping,lip lock,french kiss are damn good.
good quality even 0.2 pixel.
but very slow.need lighting lora.but i afraid of lighting lora will spoil the model.
overall opinion. perfect model for nsfw video. thank you guys.
update.
i notice bluring in teeth for all resolution.i dont know why.
controling body part via prompt is very good.
please use offcial prompt structure.its really work well.
before light lora its minimax h3 after lighting lora it became ltx
The male genitals were coming out looking really weirdβis there a way to fix that? I'm curious about how to write the prompt.
The male genitals were coming out looking really weirdβis there a way to fix that? I'm curious about how to write the prompt.
lora is the only option
A100 80 gpu ram with sage attention. 0.5 pixel,10s,20 steps
[INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 32427MB Staged. 0 patches attached. Force pre-loaded 210 weights: 1142 KB.
95%|ββββββββββ| 19/20 [03:40<00:16, 16.62s/it]
wan 2.2 7/10 ,
ltx 2.3 5/10,
min max h3 8/10
i dont know why blur coming.
wtf hugging face is wild