new video export

app.py

@@ -1,16 +1,40 @@
 import gradio as gr
+import cv2
+import numpy as np
 import torch
 from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
-from diffusers.utils import export_to_video
 
 pipe = DiffusionPipeline.from_pretrained("cerspense/zeroscope_v2_576w", torch_dtype=torch.float16)
 pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
 pipe.enable_model_cpu_offload()
 
+def export_to_video(video_frames):
+    # Convert the nested list to a NumPy array and scale values to 0-255 range
+    video_frames = np.array(video_frames)
+    video_frames = (video_frames * 255).astype(np.uint8)
+    
+    # Get the dimensions of the frames
+    height, width, channels = video_frames.shape[2:]
+    
+    # Define the video writer object
+    fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # Codec for .mp4 files
+    fps = 24  # Frames per second
+    video_writer = cv2.VideoWriter('output_video.mp4', fourcc, fps, (width, height))
+    
+    # Write each frame to the video
+    for i in range(video_frames.shape[0]):
+        frame = video_frames[i]
+        video_writer.write(frame)
+    
+    # Release the video writer object
+    video_writer.release()
+    
+    print("Video has been created successfully.")
+    return 'output_video.mp4'
+
 def infer(prompt):
     negative_prompt = "text, watermark, copyright, blurry, nsfw"
     video_frames = pipe(prompt, negative_prompt=negative_prompt, num_inference_steps=40, height=320, width=576, num_frames=24).frames
-    print(f"VIDEO FRAMES: {video_frames}")
     video_path = export_to_video(video_frames)
     print(video_path)
     return video_path