time tracking

app.py

@@ -5,6 +5,7 @@ import torch
 from pyctcdecode import BeamSearchDecoderCTC
 import torch
 import librosa
+import time
 
 
 lmID = "aware-ai/german-lowercase-wiki-4gram"
@@ -36,12 +37,21 @@ def translate(src, tgt, text):
 
 def transcribe(audio):
     sampling_rate = 16000
+    start_time = time.time()
     audio, sr = librosa.load(audio, sr=sampling_rate)
+    print("--- %s seconds audio loading ---" % (time.time() - start_time))
+    start_time = time.time()
     speech_timestamps = get_speech_timestamps(audio, vadmodel, sampling_rate=sampling_rate)
+    print("--- %s seconds audio timestamps---" % (time.time() - start_time))
+    start_time = time.time()
     chunks = [audio[i["start"]:i["end"]] for i in speech_timestamps]
+    print("--- %s seconds audio chunking---" % (time.time() - start_time))
+    start_time = time.time()
     transcribed = " ".join([text["text"] for text in p(chunks, chunk_length_s=20, stride_length_s=(0, 0))])
-    
+    print("--- %s seconds audio transcription ---" % (time.time() - start_time))
+    start_time = time.time()    
     punctuated = ttp(transcribed, max_length = 512)[0]["generated_text"]
+    print("--- %s seconds audio formatting ---" % (time.time() - start_time))
     
     return transcribed, punctuated