Update app.py

app.py

@@ -1,49 +1,33 @@
+from tts_voice import tts_order_voice
+import edge_tts
 import gradio as gr
-from TTS.api import TTS
 import tempfile
-import os
-import sys
+import anyio
+import asyncio
 
-# --- Model Loading (Runs only once at startup) ---
-# NOTE: Switched to a Tacotron2/LJSPEECH model, which typically uses the 
-# Python dependency 'gruut' instead of the system package 'espeak-ng'.
-try:
-    # Initialize TTS with the chosen model.
-    # This model is known to be stable and CPU-compatible.
-    # The '.to("cpu")' ensures it runs on the free hardware tier.
-    tts_model = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC", progress_bar=False).to("cpu")
-    # For Tacotron2, we typically use the name of the dataset as the speaker
-    VOICES = ["ljspeech"] 
-    DEFAULT_VOICE = VOICES[0]
-except Exception as e:
-    tts_model = None
-    VOICES = ["Model Load Error"]
-    DEFAULT_VOICE = VOICES[0]
-    # Print the error to the logs, but continue with the Gradio interface
-    print(f"Error loading TTS model (likely gruut failure or missing dependency): {e}", file=sys.stderr)
+# --- Voice Data ---
+# Uses the dictionary loaded from your tts_voice.py file
+language_dict = tts_order_voice
+DEFAULT_LANGUAGE = list(language_dict.keys())[0]
 
-
-# --- Core TTS Function (Synchronous) ---
-def text_to_speech_coqui(text, speaker_name):
+# --- Core TTS Function (Asynchronous) ---
+async def text_to_speech_edge(text, language_code):
     # 1. Input Validation
-    if not tts_model:
-        return "ERROR: TTS Model failed to load at startup. Check Space logs.", None
     if not text or not text.strip():
         return "ERROR: Input text cannot be empty.", None
 
     try:
+        voice = language_dict[language_code]
+        communicate = edge_tts.Communicate(text, voice)
+        
         # Create a temporary file path
-        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_file:
             tmp_path = tmp_file.name
 
-        # Generate audio file (Coqui TTS method)
-        tts_model.tts_to_file(
-            text=text, 
-            speaker=speaker_name, 
-            file_path=tmp_path
-        )
+        # Generate audio file
+        await communicate.save(tmp_path)
 
-        return "Speech synthesis complete: {}".format(text), tmp_path
+        return f"语音合成完成：{text}", tmp_path
 
     except Exception as e:
         # Handle all generation errors
@@ -51,25 +35,27 @@ def text_to_speech_coqui(text, speaker_name):
 
 
 # --- Gradio UI Definition ---
-input_text = gr.Textbox(lines=5, label="Input Text") 
-output_text = gr.Textbox(label="Output Text") 
-output_audio = gr.Audio(type="filepath", label="Generated Audio File") 
+input_text = gr.Textbox(lines=5, label="输入文本") 
+output_text = gr.Textbox(label="输出文本") 
+output_audio = gr.Audio(type="filepath", label="导出文件") 
 
 language = gr.Dropdown(
-    choices=VOICES, 
-    value=DEFAULT_VOICE, 
-    label="Speaker/Voice"
+    choices=list(language_dict.keys()), 
+    value=DEFAULT_LANGUAGE, 
+    label="语言"
 ) 
 
 
 # --- Gradio Interface Definition and Launch ---
+# Note: Since the core function is async, we use asyncio for launch
 interface = gr.Interface(
-    fn=text_to_speech_coqui,
+    fn=text_to_speech_edge,
     inputs=[input_text, language],
     outputs=[output_text, output_audio],
-    title="Coqui TTS (Tacotron2/CPU Optimized)",
-    description="Customizable, high-quality Text-to-Speech using a model that avoids system dependencies."
+    title="Edge TTS 文字转语音 (Cloud API)",
+    description="使用微软 Edge TTS 提供的云服务，稳定且无需本地模型加载。"
 )
 
 if __name__ == "__main__":
-    interface.launch()
+    # Ensure anyio is used to run the async Gradio app
+    anyio.run(interface.launch, backend="asyncio")