Text-to-Video
Diffusers
Safetensors
MiniMax H3
video
audio
text-to-audio-video
distillation
dmd2
few-step
fastvideo
fasth3
Instructions to use FastVideo/FastVideo-FastH3-8-Step-V2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use FastVideo/FastVideo-FastH3-8-Step-V2 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("FastVideo/FastVideo-FastH3-8-Step-V2", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
REF2VA
#2
by mkay76 - opened
Please add REF2VA. Text to VA is kind of pointless, you cant build anything serious with it.
Any news?
So this current version works on reference as far as I can tell and have tested in a crude implementation. As long as you stay to small resolutions. The issue is the only transfer out of this release is those gates and timestep which are bound up in this fl2va release. If they would just add the gates to the reference models timestep it would be fine, the actual turbo requires no adjustment. There's nothing else they have to do really as far as I can tell. The actual turbo and 10/3 shift targets and VSA work good on reference and actually help fix the distant face issues.
