Update audio2text/a2t.py

audio2text/a2t.py

@@ -2,7 +2,7 @@ import numpy as np
 
 import librosa
 import torch
-from .init import processor, model
+from .init import pipe
 
 LIMIT = 90 # limit 90 seconds
 
@@ -37,6 +37,14 @@ class A2T:
         except Exception as e:
             print("Preprocces error", e)
             return None
+
+    def __transcribe(self, inputs, task: str = None):
+        if inputs is  None:
+            print("Inputs None")
+
+        transcribed_text = pipe(inputs, generate_kwargs={"task": "transcribe"}, return_timestamps=True)["text"]
+        return transcribed_text
+            
         
     def predict(self):
         if self.mic is not None:
@@ -46,12 +54,12 @@ class A2T:
             return "please provide audio"
     
         try:
-            forced_decoder_ids = processor.get_decoder_prompt_ids(language="english", task="transcribe")
-            # audio = self.__preprocces(audio=audio, frame_rate=frame_rate)
-            inputs = processor(audio=audio, sampling_rate=16000, return_tensors="pt")
-            predicted_ids = model.generate(**inputs, max_length=400, forced_decoder_ids=forced_decoder_ids)
-            transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
-            return transcription[0]
+            # forced_decoder_ids = processor.get_decoder_prompt_ids(language="english", task="transcribe")
+            # # audio = self.__preprocces(audio=audio, frame_rate=frame_rate)
+            # inputs = processor(audio=audio, sampling_rate=16000, return_tensors="pt")
+            # predicted_ids = model.generate(**inputs, max_length=400, forced_decoder_ids=forced_decoder_ids)
+            # transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
+            return __transcribe(inputs=audio)
         except Exception as e:
             print("Predict error", e)
             return "Oops some kinda error"