codejin/diffsingerkr

1import streamlit as st2 3from Inference import Inferencer4 5def app_diffsingerkr():6    if not 'diffsingerkr_duration' in st.session_state.keys():7        st.session_state.diffsingerkr_duration = ''8    if not 'diffsingerkr_lyric' in st.session_state.keys():9        st.session_state.diffsingerkr_lyric = ''10    if not 'diffsingerkr_note' in st.session_state.keys():11        st.session_state.diffsingerkr_note = ''12    if not 'inferencer' in st.session_state.keys():13        st.session_state.inferencer = Inferencer(14            hp_path= 'Hyper_Parameters.yaml',15            checkpoint_path= 'Checkpoint/S_200000.pt',16            batch_size= 117            )18 19    st.title('DiffSinger-KR')20    st.markdown('* This code is an implementation of DiffSinger for Korean.')21    st.markdown('* When music score which is note, duration, and lyric information are entered, singing voices are synthesized accordingly.')22    st.markdown('* Due to the range of the trained dataset, the supported notes are between 65 and 89.')23    st.markdown('* Please refer to the [here](https://github.com/CODEJIN/DiffSingerKR) for the source code for training the model.')24 25    st.markdown('''---''')26    status_indicator = st.empty()27    status_indicator.header('Insert the music!')28    st.markdown('''---''')    29    example1_col, example2_col, example3_col, _ = st.columns(4)30    if example1_col.button('Example 1'):31        st.session_state.diffsingerkr_duration = '0.52,0.17,0.35,0.35,0.35,0.35,0.70,0.35,0.35,0.70,0.35,0.35,0.70,0.52,0.17,0.35,0.35,0.35,0.35,0.70,0.35,0.35,0.35,0.35,1.39'32        st.session_state.diffsingerkr_lyric = '떴,다,떴,다,비,행,기,날,아,라,날,아,라,높,이,높,이,날,아,라,우,리,비,행,기'33        st.session_state.diffsingerkr_note = '76,74,72,74,76,76,76,74,74,74,76,79,79,76,74,72,74,76,76,76,74,74,76,74,72'34        st.experimental_rerun()35    if example2_col.button('Example 2'):36        st.session_state.diffsingerkr_duration = '0.53,0.52,0.50,0.57,0.58,0.46,0.48,0.50,0.37,0.13,0.43,0.21,0.57,0.43,0.49,1.44,0.26,0.49,0.14,0.13,0.57,0.26,0.06,0.15,0.63,0.26,0.51,0.20,0.48,0.72,0.22'37        st.session_state.diffsingerkr_lyric = '만,나,고,<X>,난,외,로,움,을,<X>,알,았,어,내,겐,<X>,관,심,조,<X>,차,<X>,없,<X>,다,는,걸,<X>,알,면,서'38        st.session_state.diffsingerkr_note = '76,78,79,0,71,74,72,71,72,0,71,69,69,71,74,0,79,78,79,0,71,0,74,0,74,72,72,0,71,71,69'39        st.experimental_rerun()40    if example3_col.button('Example 3'):41        st.session_state.diffsingerkr_duration = '0.33,0.16,0.33,0.49,0.33,0.16,0.81,0.33,0.16,0.16,0.33,0.16,0.49,0.16,0.82,0.33,0.16,0.33,0.49,0.33,0.16,0.33,0.49,0.33,0.33,0.16,0.33,1.47,0.33,0.16,0.33,0.49,0.33,0.16,0.81,0.33,0.16,0.16,0.33,0.16,0.49,0.16,0.82,0.33,0.16,0.33,0.16,0.33,0.49,0.16,0.33,0.33,0.33,0.33,0.16,0.33,0.82'42        st.session_state.diffsingerkr_lyric = '마,음,울,적,한,날,에,<X>,거,리,를,걸,어,보,고,향,기,로,운,칵,테,일,에,취,해,도,보,고,한,편,의,시,가,있,는,<X>,전,시,회,장,도,가,고,밤,새,도,<X>,록,그,리,움,에,편,질,쓰,고,파'43        st.session_state.diffsingerkr_note = '80,80,80,87,85,84,82,0,84,84,84,85,84,79,79,77,77,77,80,80,78,77,75,77,80,79,80,82,80,80,80,87,85,84,82,0,84,84,84,85,84,79,79,77,77,77,79,80,80,77,75,75,77,80,79,82,80'44        st.experimental_rerun()45    st.markdown('''---''')46    duration = st.text_input('Duration', value= st.session_state.diffsingerkr_duration)47    lyric = st.text_input('Lyric', value= st.session_state.diffsingerkr_lyric)48    note = st.text_input('Note', value= st.session_state.diffsingerkr_note)49    singer = 'CSD'50    genre = 'Children'51    key_adjustment = st.select_slider(52        label= 'Key adjustment',53        options= [x for x in range(-6, 7)],54        value= 055        )56 57    if st.button("Generate!"):58        if duration != '' and lyric != '' and note != '':59            status_indicator.header('Generating...')60            audio = st.session_state.inferencer.Inference_Epoch(61                message_times_list= [[float(x) for x in duration.strip().split(',')]],62                lyrics= [[x for x in lyric.strip().split(',')]],63                notes= [[64                    (int(x) + key_adjustment if int(x) != 0 else int(x))65                    for x in note.strip().split(',')66                    ]],67                singers= [singer],68                genres= [genre]69                )[0]70 71            st.audio(72                audio,73                format="audio/wav",74                start_time=0,75                sample_rate= st.session_state.inferencer.hp.Sound.Sample_Rate76                )77 78            status_indicator.header('Done.')79 80if __name__ == '__main__':81    app_diffsingerkr()