update

main.py

@@ -54,7 +54,7 @@ def run_nx_vad(filename: str, silence_time: float = 0.3, longest_activate: float
     pattern = "(\\d+)[\r\n]VadFlagPrepare[\r\n](?:\\d+)[\r\n]VadFlagSpeaking(?:[\r\n](?:\\d+)[\r\n]VadFlagPause[\r\n](?:\\d+)[\r\n]VadFlagSpeaking)?[\r\n](\\d+)[\r\n]VadFlagNoSpeech"
 
     vad_timestamps = re.findall(pattern, raw_vad_result, flags=re.DOTALL)
-    vad_timestamps = [(float(start), float(end)) for start, end in vad_timestamps]
+    vad_timestamps = [(int(start), int(end)) for start, end in vad_timestamps]
 
     vad_timestamps: str = json.dumps(vad_timestamps, ensure_ascii=False, indent=2)
 
@@ -74,7 +74,7 @@ def run_silero_vad(filename: str, silence_time: float = 0.3, longest_activate: f
     pattern = "speech starts at (.+?)s[\r\n].*?speech ends at (.+?)s"
 
     vad_timestamps = re.findall(pattern, raw_vad_result, flags=re.DOTALL)
-    vad_timestamps = [(float(start), float(end)) for start, end in vad_timestamps]
+    vad_timestamps = [(int(float(start) * 1000), float(end) * 1000) for start, end in vad_timestamps]
 
     vad_timestamps: str = json.dumps(vad_timestamps, ensure_ascii=False, indent=2)