Update audio2text/a2t.py

audio2text/a2t.py

@@ -3,7 +3,7 @@ import numpy as np
 from .init import pipe
 
 TASK = "transcribe"
-BATCH_SIZE = 16
+BATCH_SIZE = 8
 LIMIT = 60
 
 class A2T:
@@ -14,13 +14,13 @@ class A2T:
         if inputs is  None:
             print("Inputs None")
 
-        transcribed_text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task, "language": "english"})["text"]
+        transcribed_text = pipe(inputs, batch_size=BATCH_SIZE,)["text"]
         return transcribed_text
 
     def __preprocces(self, raw: np.ndarray, sampling_rate: int):
-        chunk = raw.astype(np.float32) / 32768.0
+        chunk = raw.astype(np.float32) / 32678.0
 
-        if sampling_rate > 16000:
+        if sampling_rate != 16000:
             chunk = librosa.resample(chunk, orig_sr=sampling_rate, target_sr=16000)
 
         # chunk = chunk[:16000*LIMIT]
@@ -39,7 +39,6 @@ class A2T:
                 raise Exception("please provide audio")
 
             if isinstance(audio , np.ndarray):
-                # inputs = {"sampling_rate": 16000, "raw": audio}
                 return self.__transcribe(inputs=audio, task=TASK)
             else:
                 raise Exception("Audio is not np array")