Update app.py

app.py

@@ -3,6 +3,7 @@ import torch
 import gradio as gr
 import librosa
 import numpy as np
+import time
 
 p = pipeline("automatic-speech-recognition", model="aware-ai/wav2vec2-base-german")
 
@@ -19,24 +20,26 @@ def is_speech(wav, sr):
 
     return len(speech_timestamps) > 0
 
-def transcribe(audio, state={"text": "", "temp_text": "", "audio": None}):
+def transcribe(audio, state={"text": "", "temp_text": "", "audio": ""}):
     if state is None:
-        state={"text": "", "temp_text": "", "audio": None}
+        state={"text": "", "temp_text": "", "audio": ""}
     wav_data, _sr = librosa.load(audio, sr=16000)
     speech = is_speech(wav_data, _sr)
     if(speech):
-        if(state["audio"] is None):
+        if(state["audio"] is ""):
             state["audio"] = wav_data
         else:
             state["audio"] = np.concatenate((state["audio"], wav_data))
-
-        text = p(state["audio"])["text"] + "\n"
-        state["temp_text"] = text
     else:
+        if(state["audio"] is not ""):
+            text = p(state["audio"])["text"] + "\n"
+            state["temp_text"] = text
+        
         state["text"] += state["temp_text"]
         state["temp_text"] = ""
-        state["audio"] = None
+        state["audio"] = ""
 
+    time.sleep(0.5)
     return f'{state["text"]} ( {state["temp_text"]} )', state
 
 gr.Interface(