Update app.py

app.py

@@ -1,5 +1,5 @@
 import streamlit as st
-import whisper.audio as iaudio
+import whisper 
 import ffmpeg
 import os
 from transformers import pipeline
@@ -7,6 +7,8 @@ from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
 import numpy as np
 SAMPLE_RATE = 16000
 
+
+
 def load_audio(file: str, sr: int = SAMPLE_RATE):
 
    
@@ -23,9 +25,10 @@ def load_audio(file: str, sr: int = SAMPLE_RATE):
 if 'processor' not in locals():
 
   with st.spinner('Wait for it...'):
-
+   
    processor = AutoProcessor.from_pretrained("openai/whisper-tiny")
-   model = AutoModelForSpeechSeq2Seq.from_pretrained("openai/whisper-tiny")
+   model=whisper.load_model("tiny")
+   
 
 
 
@@ -42,7 +45,9 @@ if wav_up is not None:
       with open(wav_up.name,"wb") as f: 
         f.write(wav_up.getbuffer())         
         st.success("Saved File")
-      audio = iaudio.load_audio(wav_up.name )
+        audio = whisper.load_audio(wav_up.name)
+        audio = whisper.pad_or_trim(audio)
+       
       st.audio(wav_up.name, format="audio/wav", start_time=0)
 if st.button('Processa'):
  if wav_up is not None:
@@ -57,10 +62,13 @@ if st.button('Processa'):
         #processado=np.frombuffer(wav_up.getbuffer(), np.int16).flatten().astype(np.float32) / 32768.0
         input_features = processor(audio , return_tensors="pt").input_features 
         forced_decoder_ids = processor.get_decoder_prompt_ids(language = None, task = "transcribe")
-      
+        transcription=model.transcribe(
+            audio,
+            language = 'pt'
+        )
         predicted_ids = model.generate(input_features, forced_decoder_ids = forced_decoder_ids)
         
-        transcription = processor.batch_decode(predicted_ids, skip_special_tokens = True)
+        #transcription = processor.batch_decode(predicted_ids, skip_special_tokens = True)
         string1=''
        #  for i, segment in enumerate(transcription, start=1):
        #  write srt lines