Update app.py

app.py

@@ -3,6 +3,7 @@ import os
 import spaces
 import torch
 from transformers import pipeline, WhisperTokenizer
+import torchaudio
 import gradio as gr
 # Please note that the below import will override whisper LANGUAGES to add bambara
 # this is not the best way to do it but at least it works. for more info check the bambara_utils code
@@ -26,6 +27,25 @@ tokenizer = WhisperTokenizer.from_pretrained(model_checkpoint, language=language
 pipe = pipeline(model=model_checkpoint, tokenizer=tokenizer, device=device, revision=revision)
 
 
+def resample_audio(audio_path, target_sample_rate=16000):
+    """
+    Converts the audio file to the target sampling rate (16000 Hz).
+    
+    Args:
+        audio_path (str): Path to the audio file.
+        target_sample_rate (int): The desired sample rate.
+
+    Returns:
+        A tensor containing the resampled audio data and the target sample rate.
+    """
+    waveform, original_sample_rate = torchaudio.load(audio_path)
+    
+    if original_sample_rate != target_sample_rate:
+        resampler = torchaudio.transforms.Resample(orig_freq=original_sample_rate, new_freq=target_sample_rate)
+        waveform = resampler(waveform)
+    
+    return waveform, target_sample_rate
+
 @spaces.GPU()
 def transcribe(audio):
     """
@@ -37,8 +57,12 @@ def transcribe(audio):
     Returns:
         A string representing the transcribed text.
     """
+    # Convert the audio to 16000 Hz
+    waveform, sample_rate = resample_audio(audio)
+    
     # Use the pipeline to perform transcription
-    text = pipe(audio)["text"]
+    text = pipe({"array": waveform.squeeze().numpy(), "sampling_rate": sample_rate})["text"]
+    
     return text
 
 def get_wav_files(directory):