app.py

import time
import torch
import string
from espnet_model_zoo.downloader import ModelDownloader
from espnet2.bin.asr_inference import Speech2Text


import soundfile
import librosa.display
import matplotlib.pyplot as plt
import gradio as gr

lang = 'multilingual'
fs = 16000 
tag = 'ftshijt/open_li52_asr_train_asr_raw_bpe7000_valid.acc.ave_10best' 

d = ModelDownloader()
speech2text = Speech2Text(
    **d.download_and_unpack(tag),
    device="cpu",
    minlenratio=0.0,
    maxlenratio=0.0,
    ctc_weight=0.3,
    beam_size=10,
    batch_size=0,
    nbest=1
)

def text_normalizer(text):
    text = text.upper()
    return text.translate(str.maketrans('', '', string.punctuation))

def inference(audio):
  speech, rate = soundfile.read(audio.name)
  assert rate == fs, "mismatch in sampling rate"
  nbests = speech2text(speech)
  text, *_ = nbests[0]

  print(f"Input Speech: {file_name}")
  return f"ASR hypothesis: {text_normalizer(text)}"
  
inputs = gr.inputs.Audio(label="Input Audio", type="file")
outputs =  gr.outputs.Textbox(label="Output Text")

title = "wav2vec 2.0"
description = "demo for Facebook AI wav2vec 2.0 using Hugging Face transformers. To use it, simply upload your audio, or click one of the examples to load them. Read more at the links below."
article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2006.11477'>wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations</a> | <a href='https://github.com/pytorch/fairseq'>Github Repo</a> | <a href='https://huggingface.co/facebook/wav2vec2-base-960h'>Hugging Face model</a></p>"


gr.Interface(inference, inputs, outputs, title=title, description=description, article=article).launch()