Update app.py

app.py

@@ -3,33 +3,24 @@ import cv2
 import numpy as np
 import torch
 from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
-from moviepy.editor import ImageSequenceClip
 
 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)
-    
-    # Convert frames from (N, H, W, C) to list of (H, W, C)
-    frames_list = [frame for frame in video_frames]
-    
-    # Create a video clip from the frames
-    fps = 24  # Frames per second
-    clip = ImageSequenceClip(frames_list, fps=fps)
-    
-    # Write the video file
-    clip.write_videofile("output_video.mp4", codec="libx264")
-    
-    print("Video has been created successfully.")
-    return 'output_video.mp4'
+def export_to_video(frames: np.ndarray, fps: int) -> str:
+    frames = np.clip((frames * 255), 0, 255).astype(np.uint8)
+    out_file = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False)
+    writer = imageio.get_writer(out_file.name, format="FFMPEG", fps=fps)
+    for frame in frames:
+        writer.append_data(frame)
+    writer.close()
+    return out_file.name
 
 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
+    video_frames = pipe(prompt, negative_prompt=negative_prompt, num_inference_steps=40, height=320, width=576, num_frames=24).frames[0]
+   
     video_path = export_to_video(video_frames)
     print(video_path)
     return video_path