Modelpublic

ai-forever/kandinsky-video-motion-predictor

sourceHugging Faceupdated 1y agoView on Hugging Face

2likes98downloads

Model Card

Code: https://github.com/gen-ai-team/kandinsky-video-tools

Demo:

python

from models import VideoMotionPredictor

video_path = "../assets/examples/mixed/girl-listens-to-music-and-dances-happily[mixkit].mp4"

model = VideoMotionPredictor.from_pretrained("ai-forever/kandinsky-video-motion-predictor").to("cuda")
pred_dict = model.inference(video_path, return_dict=True)