add ZeroGPU decorations

webgui.py

@@ -4,6 +4,7 @@
 webui
 '''
 
+import spaces
 import os
 import random
 from datetime import datetime
@@ -252,6 +253,36 @@ def process_video(uploaded_img, uploaded_audio, width, height, length, seed, fac
 
     return final_output_path
 
+def trim_audio(file_path, output_path, max_duration=10):
+    # Load the audio file
+    audio = AudioSegment.from_wav(file_path)
+    
+    # Convert max duration to milliseconds
+    max_duration_ms = max_duration * 1000
+    
+    # Trim the audio if it's longer than max_duration
+    if len(audio) > max_duration_ms:
+        audio = audio[:max_duration_ms]
+    
+    # Export the trimmed audio
+    audio.export(output_path, format="wav")
+    print(f"Audio trimmed and saved as {output_path}")
+    return output_path
+
+@spaces.GPU(duration=300)
+def generate_video(uploaded_img, uploaded_audio, width, height, length, seed, facemask_dilation_ratio, facecrop_dilation_ratio, context_frames, context_overlap, cfg, steps, sample_rate, fps, device, progress=gr.Progress(track_tqdm=True)):
+    if is_shared_ui:
+        gr.Info("Trimming audio to max 10 seconds. Duplicate the space for unlimited audio length.")
+        uploaded_audio = trim_audio(uploaded_audio, "trimmed_audio.wav")
+
+        
+            
+    final_output_path = process_video(
+        uploaded_img, uploaded_audio, width, height, length, seed, facemask_dilation_ratio, facecrop_dilation_ratio, context_frames, context_overlap, cfg, steps, sample_rate, fps, device
+    )        
+    output_video= final_output_path
+    return final_output_path
+
 def get_maskGCT_TTS(prompt_audio_maskGCT, audio_to_clone):
     try:
         client = Client("amphion/maskgct")
@@ -356,36 +387,7 @@ with gr.Blocks() as demo:
                     <img src="https://huggingface.co/datasets/huggingface/badges/resolve/main/follow-me-on-HF-xl-dark.svg" alt="Follow me on HF">
                 </a>
             </div>
-            """)
-
-    def trim_audio(file_path, output_path, max_duration=10):
-        # Load the audio file
-        audio = AudioSegment.from_wav(file_path)
-    
-        # Convert max duration to milliseconds
-        max_duration_ms = max_duration * 1000
-    
-        # Trim the audio if it's longer than max_duration
-        if len(audio) > max_duration_ms:
-            audio = audio[:max_duration_ms]
-    
-        # Export the trimmed audio
-        audio.export(output_path, format="wav")
-        print(f"Audio trimmed and saved as {output_path}")
-        return output_path
-
-    def generate_video(uploaded_img, uploaded_audio, width, height, length, seed, facemask_dilation_ratio, facecrop_dilation_ratio, context_frames, context_overlap, cfg, steps, sample_rate, fps, device, progress=gr.Progress(track_tqdm=True)):
-        if is_shared_ui:
-            gr.Info("Trimming audio to max 10 seconds. Duplicate the space for unlimited audio length.")
-            uploaded_audio = trim_audio(uploaded_audio, "trimmed_audio.wav")
-
-        
-            
-        final_output_path = process_video(
-            uploaded_img, uploaded_audio, width, height, length, seed, facemask_dilation_ratio, facecrop_dilation_ratio, context_frames, context_overlap, cfg, steps, sample_rate, fps, device
-        )        
-        output_video= final_output_path
-        return final_output_path
+            """)    
 
     gen_maskGCT_voice_btn.click(
         fn = get_maskGCT_TTS,