app.py

from math import log2, pow
import os

import numpy as np
from scipy.fftpack import fft

import gradio as gr

from funasr import AutoModel
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks
from modelscope import snapshot_download

A4 = 440
C0 = A4 * pow(2, -4.75)
name = ["C", "C#", "D", "D#", "E", "F", "F#", "G", "G#", "A", "A#", "B"]


def get_pitch(freq):
    h = round(12 * log2(freq / C0))
    n = h % 12
    return name[n]

inference_pipeline = pipeline(
        task=Tasks.auto_speech_recognition,
        model='./models_from_modelscope/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
        vad_model="./models_from_modelscope/speech_fsmn_vad_zh-cn-16k-common-pytorch",
        punc_model="./models_from_modelscope/punc_ct-transformer_cn-en-common-vocab471067-large",
    )

def ASR(audio):

    param_dict = {}
    param_dict['use_timestamp'] = False

    rec_result = inference_pipeline(input=audio, params=param_dict)

    return rec_result[0]['text']


demo = gr.Interface(
    ASR,

    gr.Audio(type="filepath"),
    outputs = 'text',
    allow_flagging= 'auto',
    examples=[
        [os.path.join(os.path.dirname(__file__),"test1.mp3")],
    ],
)

if __name__ == "__main__":
    demo.launch(share=True)