Update app.py

app.py

@@ -1,29 +1,36 @@
 import os
 os.environ["NUMBA_DISABLE_CACHE"] = "1"
 import gradio as gr
-from openvoice import OpenVoice
+import os
+import torch
 
-# Load OpenVoice model (adjust to your setup)
-model = OpenVoice(language="en")
+# Add openvoice path
+import sys
+sys.path.append("openvoice")
 
-def clone_and_speak(audio, text):
-    output_path = "output.wav"
-    model.clone_voice(
-        source_audio_path=audio.name,
-        target_text=text,
-        output_path=output_path
-    )
-    return output_path
+from openvoice.api import ToneColorConverter
+from openvoice.inference import voice_conversion
+
+# Set up paths
+ckpt_converter = './checkpoints/converter'
+device = "cuda" if torch.cuda.is_available() else "cpu"
 
-with gr.Blocks() as demo:
-    gr.Markdown("# OpenVoice TTS - Hugging Face Space")
-    with gr.Row():
-        audio_input = gr.Audio(label="Upload voice to clone", type="file")
-        text_input = gr.Textbox(label="Enter text to synthesize")
-    with gr.Row():
-        generate_btn = gr.Button("Generate Audio")
-        audio_output = gr.Audio(label="Synthesized Output", type="filepath")
+converter = ToneColorConverter(f"{ckpt_converter}/config.json", device=device)
+converter.load_ckpt(f"{ckpt_converter}/converter.ckpt")
+
+def convert_voice(audio_file, text_prompt):
+    output_path = "./results/output.wav"
+    # You must clone reference audio using clone.sh or similar step in Dockerfile
+    voice_conversion(converter, audio_file.name, text_prompt, output_path, device)
+    return output_path
 
-    generate_btn.click(fn=clone_and_speak, inputs=[audio_input, text_input], outputs=audio_output)
+iface = gr.Interface(
+    fn=convert_voice,
+    inputs=[
+        gr.Audio(type="filepath", label="Input Voice (WAV)"),
+        gr.Textbox(label="Prompt (e.g., 'Speak in a cheerful tone')"),
+    ],
+    outputs=gr.Audio(label="Converted Voice")
+)
 
-demo.launch()
+iface.launch()