Update app.py

app.py

@@ -6,94 +6,65 @@ import langid
 from openvoice.api import BaseSpeakerTTS, ToneColorConverter
 import openvoice.se_extractor as se_extractor
 
-# Use environment variables or predefined paths
-CKPT_BASE_PATH = os.getenv('CHECKPOINT_PATH', './checkpoints')
+# Constants
+CKPT_BASE_PATH = "checkpoints"
 EN_SUFFIX = f"{CKPT_BASE_PATH}/base_speakers/EN"
 CONVERTER_SUFFIX = f"{CKPT_BASE_PATH}/converter"
-OUTPUT_DIR = "./outputs"
-
-# Ensure directories exist
-os.makedirs(CKPT_BASE_PATH, exist_ok=True)
+OUTPUT_DIR = "outputs/"
 os.makedirs(OUTPUT_DIR, exist_ok=True)
 
-def download_files():
-    """Centralized file download method with error handling"""
-    files_to_download = [
-        (f"{CONVERTER_SUFFIX}/checkpoint.pth", "converter/checkpoint.pth"),
-        (f"{CONVERTER_SUFFIX}/config.json", "converter/config.json"),
-        (f"{EN_SUFFIX}/checkpoint.pth", "base_speakers/EN/checkpoint.pth"),
-        (f"{EN_SUFFIX}/config.json", "base_speakers/EN/config.json"),
-        (f"{EN_SUFFIX}/en_default_se.pth", "base_speakers/EN/en_default_se.pth"),
-        (f"{EN_SUFFIX}/en_style_se.pth", "base_speakers/EN/en_style_se.pth")
-    ]
-    
-    for local_path, remote_path in files_to_download:
-        try:
-            os.makedirs(os.path.dirname(local_path), exist_ok=True)
-            hf_hub_download(
-                repo_id="myshell-ai/OpenVoice", 
-                filename=remote_path, 
-                local_dir=CKPT_BASE_PATH
-            )
-        except Exception as e:
-            print(f"Error downloading {remote_path}: {e}")
-            raise
+# Download necessary files
+def download_from_hf_hub(filename, local_dir="./"):
+    os.makedirs(local_dir, exist_ok=True)
+    hf_hub_download(repo_id="myshell-ai/OpenVoice", filename=filename, local_dir=local_dir)
 
-# Download files early
-download_files()
+for file in [f"{CONVERTER_SUFFIX}/checkpoint.pth", f"{CONVERTER_SUFFIX}/config.json",
+             f"{EN_SUFFIX}/checkpoint.pth", f"{EN_SUFFIX}/config.json",
+             f"{EN_SUFFIX}/en_default_se.pth", f"{EN_SUFFIX}/en_style_se.pth"]:
+    download_from_hf_hub(file)
 
-# Model Initialization with Error Handling
-try:
-    pt_device = "cpu"  # Explicitly use CPU for Hugging Face deployment
-    en_base_speaker_tts = BaseSpeakerTTS(f"{EN_SUFFIX}/config.json", device=pt_device)
-    en_base_speaker_tts.load_ckpt(f"{EN_SUFFIX}/checkpoint.pth")
+# Initialize models
+pt_device = "cpu"
+en_base_speaker_tts = BaseSpeakerTTS(f"{EN_SUFFIX}/config.json", device=pt_device)
+en_base_speaker_tts.load_ckpt(f"{EN_SUFFIX}/checkpoint.pth")
 
-    tone_color_converter = ToneColorConverter(f"{CONVERTER_SUFFIX}/config.json", device=pt_device)
-    tone_color_converter.load_ckpt(f"{CONVERTER_SUFFIX}/checkpoint.pth")
+tone_color_converter = ToneColorConverter(f"{CONVERTER_SUFFIX}/config.json", device=pt_device)
+tone_color_converter.load_ckpt(f"{CONVERTER_SUFFIX}/checkpoint.pth")
 
-    en_source_default_se = torch.load(f"{EN_SUFFIX}/en_default_se.pth")
-    en_source_style_se = torch.load(f"{EN_SUFFIX}/en_style_se.pth")
-except Exception as model_init_error:
-    print(f"Model initialization error: {model_init_error}")
-    raise
+en_source_default_se = torch.load(f"{EN_SUFFIX}/en_default_se.pth")
+en_source_style_se = torch.load(f"{EN_SUFFIX}/en_style_se.pth")
 
+# Main prediction function
 def predict(prompt, style, audio_file_pth, tau):
     if len(prompt) < 2 or len(prompt) > 200:
         return "Text should be between 2 and 200 characters.", None
 
     try:
-        target_se, _ = se_extractor.get_se(
-            audio_file_pth, 
-            tone_color_converter, 
-            target_dir=OUTPUT_DIR, 
-            vad=True
-        )
+        target_se, _ = se_extractor.get_se(audio_file_pth, tone_color_converter, target_dir=OUTPUT_DIR, vad=True)
     except Exception as e:
-        return f"Error extracting tone: {str(e)}", None
+        return f"Error getting target tone color: {str(e)}", None
 
-    try:
-        src_path = f"{OUTPUT_DIR}/tmp.wav"
-        en_base_speaker_tts.tts(prompt, src_path, speaker=style, language="English")
-
-        save_path = f"{OUTPUT_DIR}/output.wav"
-        tone_color_converter.convert(
-            audio_src_path=src_path,
-            src_se=en_source_style_se if style != "default" else en_source_default_se,
-            tgt_se=target_se,
-            output_path=save_path,
-            tau=tau
-        )
+    src_path = f"{OUTPUT_DIR}/tmp.wav"
+    en_base_speaker_tts.tts(prompt, src_path, speaker=style, language="English")
+
+    save_path = f"{OUTPUT_DIR}/output.wav"
+    tone_color_converter.convert(
+        audio_src_path=src_path,
+        src_se=en_source_style_se if style != "default" else en_source_default_se,
+        tgt_se=target_se,
+        output_path=save_path,
+        tau=tau
+    )
 
-        return "Voice cloning completed successfully.", save_path
-    except Exception as conversion_error:
-        return f"Voice conversion error: {conversion_error}", None
+    return "Voice cloning completed successfully.", save_path
 
+# Gradio interface
 def create_demo():
     with gr.Blocks() as demo:
-        gr.Markdown("# OpenVoice: Instant Voice Cloning")
+        gr.Markdown("# OpenVoice: Instant Voice Cloning with fine-tuning")
         
         with gr.Row():
-            input_text = gr.Textbox(label="Text to speak", placeholder="Enter text (2-200 chars)")
+            input_text = gr.Textbox(label="Text to speak", placeholder="Enter text here (2-200 characters)")
             style = gr.Dropdown(
                 label="Style",
                 choices=["default", "whispering", "cheerful", "terrified", "angry", "sad", "friendly"],
@@ -102,13 +73,7 @@ def create_demo():
         
         with gr.Row():
             reference_audio = gr.Audio(label="Reference Audio", type="filepath")
-            tau_slider = gr.Slider(
-                minimum=0.1, 
-                maximum=1.0, 
-                value=0.7, 
-                label="Voice Similarity", 
-                info="Higher values = more similar to reference"
-            )
+            tau_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, label="Tau (Voice similarity)", info="Higher values make the output more similar to the reference voice")
 
         submit_button = gr.Button("Generate Voice")
         
@@ -123,5 +88,7 @@ def create_demo():
 
     return demo
 
-# Hugging Face Space compatibility
-demo = create_demo()
+# Launch the demo
+if __name__ == "__main__":
+    demo = create_demo()
+    demo.launch()