import torch
from diffusers import DiffusionPipeline
from diffusers.utils import export_to_video
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("Wan-AI/Wan2.1-T2V-1.3B", dtype=torch.bfloat16, device_map="cuda")
pipe.load_lora_weights("AlekseyCalvin/Wan1.3B_CausVid_LoRA_MisubiDiffusers_Conversion")
prompt = "A man with short gray hair plays a red electric guitar."
output = pipe(prompt=prompt).frames[0]
export_to_video(output, "output.mp4")CausVid Accelerator Misubi-format Conversion
Text/Image to Video Low Rank Adapter (LoRA)
For Wan2.1 1.3B family models (Base, Skyreels V2, FunINP1.3B, etc...)
Based on Kijal's original ComfyUI-oriented LoRA extraction of the CausVid accelerated mod of Wan2.1 1.3B.
This conversion might work better than Kijal's source for those using CausVid outside of a ComfyUI environment.
Try this version if: Kijal's original LoRA does not work at all for you and crashes/fails to infer.
For example: This version works in DrawThings, whereas Kijal's original does not (as of the time I write this) & simply crashes.
This might (or might not) extend to other given non-Comfy Wan-inference environments/backends/frameworks.
Please share any other confirmed use cases in the "Community" tab!
- Downloads last month
- 52
Model tree for AlekseyCalvin/Wan1.3B_CausVid_LoRA_MisubiDiffusers_Conversion
Base model
Wan-AI/Wan2.1-T2V-1.3B