Update app.py

app.py

@@ -3,7 +3,7 @@ import torch
 from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
 import gradio as gr
 import sox
-
+import scipy.signal as sps
 
 
 def convert(inputfile, outfile):
@@ -13,12 +13,20 @@ def convert(inputfile, outfile):
     )
     sox_tfm.build(inputfile, outfile)
 
+def read_file(wav):
+    sample_rate, signal = wav_file                                                                                                                        
+    signal = signal.mean(-1)                                                                                                                              
+    number_of_samples = round(len(signal) * float(16000) / sample_rate)                                                                                   
+    resampled_signal = sps.resample(signal, number_of_samples)
+    return resampled_signal
 
 
 def parse_transcription(wav_file):
-    filename = wav_file.name.split('.')[0]
+    '''filename = wav_file.name.split('.')[0]
     convert(wav_file.name, filename + "16k.wav")
     speech, _ = sf.read(filename + "16k.wav")
+    '''
+    speech = read_file(wav_file)
     input_values = processor(speech, sampling_rate=16_000, return_tensors="pt").input_values
 
     logits = model(input_values).logits
@@ -36,6 +44,8 @@ model = Wav2Vec2ForCTC.from_pretrained("Harveenchadha/vakyansh-wav2vec2-hindi-hi
 processor = Wav2Vec2Processor.from_pretrained("Harveenchadha/vakyansh-wav2vec2-hindi-him-4200")
 model = Wav2Vec2ForCTC.from_pretrained("Harveenchadha/vakyansh-wav2vec2-hindi-him-4200")
     
-input_ = gr.inputs.Audio(source="microphone", type="file") 
+#input_ = gr.inputs.Audio(source="microphone", type="file") 
+input_ = gr.inputs.Audio(source="microphone", type="numpy") 
+
 gr.Interface(parse_transcription, inputs = input_,  outputs="text", 
              analytics_enabled=False, show_tips=False, enable_queue=True).launch(inline=False);