Method preprocess_video

diffsynth/utils/__init__.py:70–74

(self, video, torch_dtype=None, device=None, pattern="B C T H W", min_value=-1, max_value=1)

    def preprocess_video(self, video, torch_dtype=None, device=None, pattern="B C T H W", min_value=-1, max_value=1):
        # Transform a list of PIL.Image to torch.Tensor
        video = [self.preprocess_image(image, torch_dtype=torch_dtype, device=device, min_value=min_value, max_value=max_value) for image in video]
        video = torch.stack(video, dim=pattern.index("T") // 2)
        return video


    def vae_output_to_image(self, vae_output, pattern="B C H W", min_value=-1, max_value=1):