Update app.py

app.py

@@ -1,6 +1,8 @@
 import re
+import os
 import ffmpy
 import asyncio
+import tarfile
 import edge_tts
 import subprocess
 import gradio as gr
@@ -9,6 +11,11 @@ from gradio_client import Client
 from http.client import RemoteDisconnected
 from list_dict import translates, speakers
 
+if not os.path.exists('pretrained_models'):
+    # 解压tar.gz文件
+    with tarfile.open('2stems.tar.gz', 'r:gz') as tar_ref:
+        tar_ref.extractall('./pretrained_models/2stems')
+
 translate = translates
 tr = list(translate.keys())[9]
 language = translate[tr]
@@ -44,8 +51,8 @@ def left_justified(audio):
     raise gr.Error('No start sound detected!')
   return start_justified
   
-def time_verify():
-  audios = [vocals_monorail, text_to_speech]
+def time_verify(vocals_audio, target_audio):
+  audios = [vocals_audio, target_audio]
   justified = []
   time_lists = []
 
@@ -134,7 +141,7 @@ def video_inputs(video, TR_LANGUAGE, LANGUAGE, SPEAKER):
     await communicate.save(text_to_speech)
   asyncio.run(amain())
 
-  j_time, r_time = time_verify()
+  r_time = time_verify(vocals_monorail, text_to_speech)
   ff = ffmpy.FFmpeg(
       inputs={
           text_to_speech: None
@@ -144,6 +151,7 @@ def video_inputs(video, TR_LANGUAGE, LANGUAGE, SPEAKER):
       }
   )
   ff.run()
+  j_time = time_verify(vocals_monorail, output_rate_audio)
   if j_time > 0:
       ff = ffmpy.FFmpeg(
           inputs={