.

options/Video_model/Model.py

@@ -7,6 +7,7 @@ from .tdd_svd_scheduler import TDDSVDStochasticIterativeScheduler
 from .utils import load_lora_weights, save_video
 from typing import Optional
 from glob import glob
+import torchvision.transforms as transforms
 
 svd_path = 'stabilityai/stable-video-diffusion-img2vid-xt-1-1'
 lora_repo_path = 'RED-AIGC/TDD'
@@ -47,14 +48,9 @@ def Video(
     base_count = len(glob(os.path.join(output_folder, "*.mp4")))
     video_path = os.path.join(output_folder, f"{base_count:06d}.mp4")
 
-    if isinstance(image, torch.Tensor):
-        # Convert BFloat16 to Float32 if necessary
-        if image.dtype == torch.bfloat16:
-            image = image.to(dtype=torch.float32)
-    elif isinstance(image, Image.Image):
-        # Convert PIL Image to a supported tensor
-        image = torch.tensor(image).float()
-
+    if isinstance(image, Image.Image):
+        transform = transforms.ToTensor()
+        image = transform(image).unsqueeze(0).to(device).float()
 
     with torch.autocast(device):
         frames = pipeline(