Adjust for ZeroGPU

.history/app_20250202161356.py

@@ -0,0 +1,34 @@
+import subprocess
+
+# Run the setup.py install command
+try:
+    subprocess.run(['python', 'setup.py', 'install', '--user'], check=True)
+    print("Installation successful.")
+except subprocess.CalledProcessError as e:
+    print(f"Installation failed with error: {e}")
+
+import gradio as gr
+import torch
+from TTS.api import TTS
+
+# Get device
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# Init TTS
+tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
+
+def voice_clone(text: str, speaker_wav: str, language: str):
+    # Run TTS
+    print("Speaker wav:", speaker_wav)
+    tts.tts_to_file(text=text, speaker_wav=speaker_wav, language=language, file_path="output.wav")
+    return "output.wav"
+
+iface = gr.Interface(fn=voice_clone, theme="Nymbo/Nymbo_Theme", 
+                     inputs=[gr.Textbox(lines=2, placeholder="Enter the text...", label="Text"),
+                             gr.Audio(type="filepath", label="Upload audio file"),
+                             gr.Radio(['ru', 'en', 'zh-cn', 'ja', 'de', 'fr', 'it', 'pt', 'pl', 'tr', 'ko', 'nl', 'cs', 'ar', 'es', 'hu'], label="language"),
+                            ],
+                     outputs=gr.Audio(type="filepath", label="Generated audio file"), 
+                     title="Voice Cloning")
+
+iface.launch()

.history/app_20250202161532.py

@@ -0,0 +1,66 @@
+import subprocess
+import spaces
+import os
+
+# Run the setup.py install command
+try:
+    subprocess.run(['python', 'setup.py', 'install', '--user'], check=True)
+    print("Installation successful.")
+except subprocess.CalledProcessError as e:
+    print(f"Installation failed with error: {e}")
+
+import gradio as gr
+import torch
+from TTS.api import TTS
+
+# Get device
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# Initialize TTS model globally but load it inside the GPU-decorated function
+tts = None
+
+@spaces.GPU(duration=120) # Voice cloning can take longer than default 60s
+def initialize_tts():
+    global tts
+    if tts is None:
+        tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
+    return tts
+
+@spaces.GPU(duration=120)
+def voice_clone(text: str, speaker_wav: str, language: str):
+    global tts
+    # Initialize TTS if not already done
+    if tts is None:
+        tts = initialize_tts()
+    
+    # Create output directory if it doesn't exist
+    os.makedirs("outputs", exist_ok=True)
+    output_path = os.path.join("outputs", "output.wav")
+    
+    # Run TTS
+    print("Speaker wav:", speaker_wav)
+    tts.tts_to_file(text=text, 
+                    speaker_wav=speaker_wav, 
+                    language=language, 
+                    file_path=output_path)
+    return output_path
+
+# Create Gradio interface
+iface = gr.Interface(
+    fn=voice_clone,
+    theme="Nymbo/Nymbo_Theme",
+    inputs=[
+        gr.Textbox(lines=2, placeholder="Enter the text...", label="Text"),
+        gr.Audio(type="filepath", label="Upload audio file"),
+        gr.Radio(
+            ['ru', 'en', 'zh-cn', 'ja', 'de', 'fr', 'it', 'pt', 'pl', 'tr', 'ko', 'nl', 'cs', 'ar', 'es', 'hu'],
+            label="language"
+        ),
+    ],
+    outputs=gr.Audio(type="filepath", label="Generated audio file"),
+    title="Voice Cloning",
+    description="Upload a voice sample and enter text to clone the voice. Processing may take 1-2 minutes."
+)
+
+# Launch with queue enabled for better handling of GPU resources
+iface.queue().launch()

app.py

@@ -1,4 +1,6 @@
 import subprocess
+import spaces
+import os
 
 # Run the setup.py install command
 try:
@@ -14,21 +16,51 @@ from TTS.api import TTS
 # Get device
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# Init TTS
-tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
+# Initialize TTS model globally but load it inside the GPU-decorated function
+tts = None
 
+@spaces.GPU(duration=120) # Voice cloning can take longer than default 60s
+def initialize_tts():
+    global tts
+    if tts is None:
+        tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
+    return tts
+
+@spaces.GPU(duration=120)
 def voice_clone(text: str, speaker_wav: str, language: str):
+    global tts
+    # Initialize TTS if not already done
+    if tts is None:
+        tts = initialize_tts()
+    
+    # Create output directory if it doesn't exist
+    os.makedirs("outputs", exist_ok=True)
+    output_path = os.path.join("outputs", "output.wav")
+    
     # Run TTS
     print("Speaker wav:", speaker_wav)
-    tts.tts_to_file(text=text, speaker_wav=speaker_wav, language=language, file_path="output.wav")
-    return "output.wav"
-
-iface = gr.Interface(fn=voice_clone, theme="Nymbo/Nymbo_Theme", 
-                     inputs=[gr.Textbox(lines=2, placeholder="Enter the text...", label="Text"),
-                             gr.Audio(type="filepath", label="Upload audio file"),
-                             gr.Radio(['ru', 'en', 'zh-cn', 'ja', 'de', 'fr', 'it', 'pt', 'pl', 'tr', 'ko', 'nl', 'cs', 'ar', 'es', 'hu'], label="language"),
-                            ],
-                     outputs=gr.Audio(type="filepath", label="Generated audio file"), 
-                     title="Voice Cloning")
-
-iface.launch()
+    tts.tts_to_file(text=text, 
+                    speaker_wav=speaker_wav, 
+                    language=language, 
+                    file_path=output_path)
+    return output_path
+
+# Create Gradio interface
+iface = gr.Interface(
+    fn=voice_clone,
+    theme="Nymbo/Nymbo_Theme",
+    inputs=[
+        gr.Textbox(lines=2, placeholder="Enter the text...", label="Text"),
+        gr.Audio(type="filepath", label="Upload audio file"),
+        gr.Radio(
+            ['ru', 'en', 'zh-cn', 'ja', 'de', 'fr', 'it', 'pt', 'pl', 'tr', 'ko', 'nl', 'cs', 'ar', 'es', 'hu'],
+            label="language"
+        ),
+    ],
+    outputs=gr.Audio(type="filepath", label="Generated audio file"),
+    title="Voice Cloning",
+    description="Upload a voice sample and enter text to clone the voice. Processing may take 1-2 minutes."
+)
+
+# Launch with queue enabled for better handling of GPU resources
+iface.queue().launch()