app.py

import streamlit as st
import requests
import json
import os
from pytube import YouTube
#import librosa
import soundfile as sf
from glob import glob
from scipy.io.wavfile import read, write
import io

os.system("apt install ffmpeg")

API_TOKEN = "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"
API_URL_ktp = "https://api-inference.huggingface.co/models/Kittipong/wav2vec2-th-vocal-domain"
API_URL_airesearch = "https://api-inference.huggingface.co/models/airesearch/wav2vec2-large-xlsr-53-th"
headers = {"Authorization": f"Bearer {API_TOKEN}"}

st.title('Auto lyric recognizer🎤')

col1, col2 = st.columns(2)

def query(API_URL,data):
    response = requests.request("POST", API_URL, headers=headers, data=data)
    return json.loads(response.content.decode("utf-8"))

def split_file(path):
    file_name = path
    audio, sr = librosa.load(file_name)

    buffer = 60 * sr
    samples_total = len(audio)
    samples_wrote = 0
    counter = 1

    while samples_wrote < samples_total:

        if buffer > (samples_total - samples_wrote):
            buffer = samples_total - samples_wrote

        block = audio[samples_wrote : (samples_wrote + buffer)]
        out_filename = "split_" + str(counter) + ".wav"

        sf.write(out_filename, block, sr)
        counter += 1
        samples_wrote += buffer

def audio_analysis(audio):
    with col2:
        if audio!=0:
            output1 = query(API_URL_ktp,audio)
            output2 = query(API_URL_airesearch,audio)
            
            with st.expander("ผลลัพท์จากโมเดล Kittipong/wav2vec2-th-vocal-domain"):
                st.write(f"{output1}")
            with st.expander("ผลลัพท์จากโมเดล airesearch/wav2vec2-large-xlsr-53-th"):
                st.write(f"{output2}")
        else:

            paths = glob('/content/split_*.wav')
            sentence1 = ""
            sentence2 = ""
            for i in range(len(paths)):
                path = f'/content/split_{i+1}.wav'
                with open(f"{path}", "rb") as wavfile:
                    input_wav = wavfile.read()
                rate, data = read(io.BytesIO(input_wav))
                bytes_wav = bytes()
                byte_io = io.BytesIO(bytes_wav)
                write(byte_io, rate, data)
                audio = byte_io.read()
                stat1,stat2 = True, True
                
                while(stat1):
                    op1 = query(API_URL_ktp,audio)
                    if 'text' in op1:
                        sentence1+=op1['text']
                        # st.text(op1)
                        stat1 = False
                    else : continue

                while(stat2):
                    op2 = query(API_URL_airesearch,audio)
                    if 'text' in op2:
                        sentence2+=op2['text']
                        # st.text(op2['text'])
                        stat2 = False
                    else : continue  

            with st.expander("ผลลัพท์จากโมเดล Kittipong/wav2vec2-th-vocal-domain"):
                st.write(f"{sentence1}")
            with st.expander("ผลลัพท์จากโมเดล airesearch/wav2vec2-large-xlsr-53-th"):
                st.write(f"{sentence2}")



class input_process:
    def audio_file():

        uploaded_file = st.file_uploader("เลือกไฟล์")
        if uploaded_file is not None:
        # To read file as bytes:
            bytes_data = uploaded_file.getvalue()
            st.audio(bytes_data,'audio/mp3')
            if st.button("ประมวลผล"):
                audio_analysis(bytes_data)

    def url():
        link = st.text_input('ใส่url', 'https://www.youtube.com/watch?v=-BOtfXCIDcs&list=RD-BOtfXCIDcs&start_radio=1')
        st.video(link)
        if st.button("ประมวลผล"):
            yt = YouTube(str(link))
            video = yt.streams.filter(only_audio=True).first()
            out_file = video.download(output_path='mp3')
            base,ext=os.path.splitext(out_file)

            os.rename(out_file,'song.mp3')

            os.system("spleeter separate -o output/ song.mp3")
            split_file('output/song/vocals.wav')
            # split_file(out_file,'/content/mp3/song.mp3')
            audio_analysis(0)



def main():
    
    with col1:
        option = st.selectbox(
            '🤟รูปแบบข้อมูลเสียงที่ต้องการนำเข้า',
            ('💿ไฟล์เสียง', '🌐url'))

        st.write('Your selected:', option)
        if option == '💿ไฟล์เสียง':
            input_process.audio_file()
        if option == '🌐url':
            os.system('rm -f /content/split_*.wav')
            input_process.url()

if __name__ == '__main__':
    main()
# %%