Add-Vishnu/Pywhisper_CPP_ASR

1import gradio as gr2import soundfile as sf3import tempfile4import shutil5import os6import librosa7import time8import numpy as np9import subprocess 10from pywhispercpp.model import Model11 12# model = Model('base.en', n_threads=6,models_dir="./Models") # Only english13# model = Model('base', n_threads=6,models_dir="./Models",language="hindi",translate=False)  # Multilingual14model = Model('medium', n_threads=6,models_dir="./Models",language="hindi",translate=False)  # Multilingual15 16def resample_to_16k(audio, orig_sr):17    y_resampled = librosa.resample(y=audio, orig_sr=orig_sr, target_sr = 16000)18    return y_resampled19 20def transcribe(audio):21    print(type(audio))22    sr,y = audio23    y = y.astype(np.float32)24    y /= np.max(np.abs(y))25    y_resampled = resample_to_16k(y, sr)26    27    28    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_audio:29        temp_audio_path = temp_audio.name30        sf.write(temp_audio_path, y_resampled, 16000)31 32    start_time_py = time.time()33    py_result = model.transcribe(f'{temp_audio_path}', n_threads=6)34    end_time_py = time.time()35    print("Py_result : ",py_result)36    print("--------------------------")37    print(f"Execution time using py: {end_time_py - start_time_py} seconds")38    output_text = ""39    for segment in py_result:40        output_text+=segment.text41    return output_text, (end_time_py - start_time_py)42 43 44 45demo = gr.Interface(46    transcribe,47    inputs = "microphone",48    # gr.Audio(sources=["microphone"]),49    outputs=[gr.Textbox(label="Py_Transcription"),gr.Textbox(label="Time taken for Transcription")],50    # examples=["./Samples/Hindi_1.mp3","./Samples/Hindi_2.mp3","./Samples/Tamil_1.mp3","./Samples/Tamil_2.mp3","./Samples/Marathi_1.mp3","./Samples/Marathi_2.mp3","./Samples/Nepal_1.mp3","./Samples/Nepal_2.mp3","./Samples/Telugu_1.wav","./Samples/Telugu_2.wav","./Samples/Malayalam_1.wav","./Samples/Malayalam_2.wav","./Samples/Gujarati_1.wav","./Samples/Gujarati_2.wav","./Samples/Bengali_1.wav","./Samples/Bengali_2.wav"]51    examples=["./Samples/Hindi_1.mp3","./Samples/Hindi_2.mp3","./Samples/Hindi_3.mp3","./Samples/Hindi_4.mp3","./Samples/Hindi_5.mp3"] # only hindi   # ,"./Samples/Tamil_1.mp3","./Samples/Tamil_2.mp3","./Samples/Marathi_1.mp3","./Samples/Marathi_2.mp3","./Samples/Nepal_1.mp3","./Samples/Nepal_2.mp3","./Samples/Telugu_1.wav","./Samples/Telugu_2.wav","./Samples/Malayalam_1.wav","./Samples/Malayalam_2.wav","./Samples/Gujarati_1.wav","./Samples/Gujarati_2.wav","./Samples/Bengali_1.wav","./Samples/Bengali_2.wav"]52)53 54demo.launch()