Update app.py

app.py

@@ -71,25 +71,55 @@ def check_audio_length(audio_path, max_duration=120):
         return False
 
 def synthesize_and_convert_voice(text, language_iso, voice_audio_path, speed):
-    tts_synthesis = TTS(model_name=f"tts_models/{language_iso}/fairseq/vits", )
+    tts_synthesis = TTS(model_name=f"tts_models/{language_iso}/fairseq/vits")
     wav_data = tts_synthesis.tts(text, speed=speed)
     tts_conversion = TTS(model_name="voice_conversion_models/multilingual/vctk/freevc24", progress_bar=False)
-    output_file = "output/docout.wav"
-    os.makedirs("output", exist_ok=True)
-    converted_audio = tts_conversion.voice_conversion_to_file(wav_data, target_wav=voice_audio_path,
-                                                              file_path=output_file)
-    return converted_audio
+
+    # Write wav_data to temporary file
+    with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as temp_tts_wav_file:
+        temp_tts_wav_path = temp_tts_wav_file.name
+        write(temp_tts_wav_path, 22050, wav_data)
+
+    # Prepare output temporary file
+    with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as temp_output_wav_file:
+        temp_output_wav_path = temp_output_wav_file.name
+
+    tts_conversion.voice_conversion_to_file(temp_tts_wav_path, target_wav=voice_audio_path,
+                                            file_path=temp_output_wav_path)
+
+    # Read converted audio from temp_output_wav_path
+    output_sample_rate, output_audio_data = read(temp_output_wav_path)
+
+    # Remove temporary files
+    os.remove(temp_tts_wav_path)
+    os.remove(temp_output_wav_path)
+
+    return (output_sample_rate, output_audio_data)
 
 def synthesize_speech(text, speaker_wav_path, language_iso, speed):
-    output_file_xtts = "output/undocout.wav"
-    tts.tts_to_file(text=text, file_path=output_file_xtts, speed=speed, speaker_wav=speaker_wav_path,
-                    language=language_iso)
+    # Generate speech using tts and save to temporary file
+    with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as temp_tts_output:
+        temp_tts_output_path = temp_tts_output.name
+        tts.tts_to_file(text=text, file_path=temp_tts_output_path, speed=speed,
+                        speaker_wav=speaker_wav_path, language=language_iso)
+
     tts_conversion = TTS(model_name="voice_conversion_models/multilingual/vctk/freevc24", progress_bar=False)
-    output_file = "output/docout.wav"
-    os.makedirs("output", exist_ok=True)
-    converted_audio = tts_conversion.voice_conversion_to_file(output_file_xtts, target_wav=speaker_wav_path,
-                                                              file_path=output_file)
-    return converted_audio
+
+    # Prepare output temporary file
+    with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as temp_output_wav_file:
+        temp_output_wav_path = temp_output_wav_file.name
+
+    tts_conversion.voice_conversion_to_file(temp_tts_output_path, target_wav=speaker_wav_path,
+                                            file_path=temp_output_wav_path)
+
+    # Read converted audio from temp_output_wav_path
+    output_sample_rate, output_audio_data = read(temp_output_wav_path)
+
+    # Remove temporary files
+    os.remove(temp_tts_output_path)
+    os.remove(temp_output_wav_path)
+
+    return (output_sample_rate, output_audio_data)
 
 def get_language_code(selected_language):
     if selected_language in language_options:
@@ -99,24 +129,37 @@ def get_language_code(selected_language):
     else:
         return None
 
-def process_speech(text, speaker_wav, selected_language, speed):
+def process_speech(text, speaker_wav_path, selected_language, speed):
     language_code = get_language_code(selected_language)
 
     if language_code is None:
         raise ValueError("Выбранный язык не поддерживается.")
 
-    # Проверка длины аудио
-    if not check_audio_length(speaker_wav):
+    if speaker_wav_path is None:
+        error_message = "Пожалуйста, загрузите аудио файл говорящего."
+        error = gr.Error(error_message, duration=5)
+        raise error
+
+    # Check audio length
+    audio = AudioSegment.from_file(speaker_wav_path)
+    duration = audio.duration_seconds
+    if duration > 120:
         error_message = "Длина аудио превышает допустимый лимит в 2 минуты."
         error = gr.Error(error_message, duration=5)
         raise error
 
-    cleaned_wav_path = clean_audio(speaker_wav)
+    # Clean audio
+    cleaned_wav_path = clean_audio(speaker_wav_path)
 
     if selected_language in other_language:
-        return synthesize_and_convert_voice(text, language_code, cleaned_wav_path, speed)
+        output_audio_data = synthesize_and_convert_voice(text, language_code, cleaned_wav_path, speed)
     else:
-        return synthesize_speech(text, cleaned_wav_path, language_code, speed)
+        output_audio_data = synthesize_speech(text, cleaned_wav_path, language_code, speed)
+
+    # Remove temporary files
+    os.remove(cleaned_wav_path)
+
+    return output_audio_data 
 
 def restart_program():
     python = sys.executable