app.py

import collections
#import datetime
#import glob
import numpy as np
#import pathlib
import pandas as pd
import pretty_midi
import seaborn as sns

from matplotlib import pyplot as plt
from typing import Optional

import tensorflow as tf

import keras

from tensorflow.keras.utils import custom_object_scope

import streamlit as st

from midi2audio import FluidSynth
import tempfile
import os
import base64

def midi_to_notes(midi_file: str) -> pd.DataFrame:
  pm = pretty_midi.PrettyMIDI(midi_file)
  instrument = pm.instruments[0]
  notes = collections.defaultdict(list)

  sorted_notes = sorted(instrument.notes, key=lambda note: note.start)
  prev_start = sorted_notes[0].start

  for note in sorted_notes:
    start = note.start
    end = note.end
    notes['pitch'].append(note.pitch)
    notes['start'].append(start)
    notes['end'].append(end)
    notes['step'].append(start - prev_start)
    notes['duration'].append(end - start)
    prev_start = start

  return pd.DataFrame({name: np.array(value) for name, value in notes.items()})

def notes_to_midi(
  notes: pd.DataFrame,
  out_file: str,
  instrument_name: str,
  velocity: int = 100,
) -> pretty_midi.PrettyMIDI:

  pm = pretty_midi.PrettyMIDI()
  instrument = pretty_midi.Instrument(
      program=pretty_midi.instrument_name_to_program(
          instrument_name))

  prev_start = 0
  for i, note in notes.iterrows():
    start = float(prev_start + note['step'])
    end = float(start + note['duration'])
    note = pretty_midi.Note(
        velocity=velocity,
        pitch=int(note['pitch']),
        start=start,
        end=end,
    )
    instrument.notes.append(note)
    prev_start = start

  pm.instruments.append(instrument)
  pm.write(out_file)
  return pm

def plot_roll(notes: pd.DataFrame, count: Optional[int] = None):
  if count:
    title = f'First {count} notes'
  else:
    title = f'Whole track'
    count = len(notes['pitch'])
  plt.figure(figsize=(20, 4))
  plot_pitch = np.stack([notes['pitch'], notes['pitch']], axis=0)
  plot_start_stop = np.stack([notes['start'], notes['end']], axis=0)
  plt.plot(
      plot_start_stop[:, :count], plot_pitch[:, :count], color="b", marker=".")
  plt.xlabel('Time [s]')
  plt.ylabel('Pitch')
  _ = plt.title(title)
  
def plot_distributions(notes: pd.DataFrame, drop_percentile=2.5):
  plt.figure(figsize=[15, 5])
  plt.subplot(1, 3, 1)
  sns.histplot(notes, x="pitch", bins=20)

  plt.subplot(1, 3, 2)
  max_step = np.percentile(notes['step'], 100 - drop_percentile)
  sns.histplot(notes, x="step", bins=np.linspace(0, max_step, 21))
  
def predict_next_note(
    notes: np.ndarray,
    model: tf.keras.Model,
    temperature: float = 1.0) -> tuple[int, float, float]:

  assert temperature > 0

  inputs = tf.expand_dims(notes, 0)

  predictions = model.predict(inputs)
  pitch_logits = predictions['pitch']
  step = predictions['step']
  duration = predictions['duration']

  pitch_logits /= temperature
  pitch = tf.random.categorical(pitch_logits, num_samples=1)
  pitch = tf.squeeze(pitch, axis=-1)
  duration = tf.squeeze(duration, axis=-1)
  step = tf.squeeze(step, axis=-1)

  step = tf.maximum(0, step)
  duration = tf.maximum(0, duration)

  return int(pitch), float(step), float(duration)

def mse_with_positive_pressure(y_true: tf.Tensor, y_pred: tf.Tensor):
  mse = (y_true - y_pred) ** 2
  positive_pressure = 10 * tf.maximum(-y_pred, 0.0)
  return tf.reduce_mean(mse + positive_pressure)

def calcular_duracion_midi(archivo_midi):
    midi = pretty_midi.PrettyMIDI(archivo_midi)
    return midi.get_end_time()
    
def main():
    seed = 42
    tf.random.set_seed(seed)
    np.random.seed(seed)

    st.title('GENERADOR DE MELODIAS CON RNN')

    
    # Rutas de archivos
    #sample_file = 'Preludes 2 Through Major keys 39.mid'
    out_file = 'output.mid'
    uploaded_file = st.file_uploader("Sube un archivo MIDI")
    
    model=''
    pesos=''
    
    option = st.selectbox(
    "Elige con que modelo entrenar",
    ("Maestro", "Lakh"))

    option_musica = st.selectbox(
    "Elige instrumento a generar las melodias",
    ("Piano", "Chromatic Percussion", "Organ", "Guitar", "Bass", "Strings", "Ensemble", "Brass",
    "Reed", "Pipe", "Synth Lead", "Synth Pad", "Synth Effects", "Ethnic", "Percussive", "Sound Effects"))

    num_predictions = st.number_input("Ingrese el número de notas:", min_value=100, max_value=150, value=120, step=1)
    
    if uploaded_file and option is not None:

        if option=="Maestro":
            model="mi_modelo_music.h5"
            pesos="mi_pesos_music.h5"
        else:
            model="mi_modelo03_music.h5"
            pesos="mi_pesos03_music.h5"
        
        st.subheader("Archivo cargado:")
        st.write(uploaded_file.name)

        st.subheader("Modelo elegido:")
        st.write(option)
        
        # Guardar el archivo en una ubicación temporal
        with open(uploaded_file.name, 'wb') as f:
            f.write(uploaded_file.getbuffer())
            
        sample_file=uploaded_file.name

        # Duracion del MIDI
        duracion = calcular_duracion_midi(sample_file)
        minutos, segundos = divmod(duracion, 60)
        
        st.write(f"La duración del archivo MIDI es: {int(minutos)} minutos y {segundos:.2f} segundos.")
        
        # Cargar modelo y pesos
        with custom_object_scope({'mse_with_positive_pressure': mse_with_positive_pressure}):
            model = keras.models.load_model(model)
    
        model.load_weights(pesos, skip_mismatch=False, by_name=False, options=None)
        
        # Convertir MIDI generado por el modelo a archivo WAV
        pm = pretty_midi.PrettyMIDI(sample_file)
        instrument_name = ""

        if option_musica is not None:
            if option_musica=="Piano":
                instrument_name="Acoustic Grand Piano"
            elif option_musica=="Chromatic Percussion":
                instrument_name="Celesta"
            elif option_musica=="Organ":
                instrument_name="Hammond Organ"
            elif option_musica=="Guitar":
                instrument_name="Acoustic Guitar (nylon)"
            elif option_musica=="Bass":
                instrument_name="Acoustic Bass"
            elif option_musica=="Strings":
                instrument_name="Violin"
            elif option_musica=="Ensemble":
                instrument_name="String Ensemble 1"
            elif option_musica=="Brass":
                instrument_name="Trumpet"
            elif option_musica=="Reed":
                instrument_name="Soprano Sax"
            elif option_musica=="Pipe":
                instrument_name="Piccolo"
            elif option_musica=="Synth Lead":
                instrument_name="Lead 2 (sawtooth)"
            elif option_musica=="Synth Pad":
                instrument_name="Pad 2 (warm)"
            elif option_musica=="Synth Effects":
                instrument_name="FX 2 (soundtrack)"
            elif option_musica=="Ethnic":
                instrument_name="Banjo"
            elif option_musica=="Percussive":
                instrument_name="Melodic Tom"
            elif option_musica=="Sound Effects":
                instrument_name="Guitar Fret Noise"
            else:
                instrument_name=pretty_midi.program_to_instrument_name(pm.instruments[0].program)
        
        raw_notes = midi_to_notes(sample_file)
        key_order = ['pitch', 'step', 'duration']
        seq_length = 25
        vocab_size = 128
        temperature = 2.0
        
        sample_notes = np.stack([raw_notes[key] for key in key_order], axis=1)
        input_notes = (sample_notes[:seq_length] / np.array([vocab_size, 1, 1]))
        generated_notes = []
        prev_start = 0
        for _ in range(num_predictions):
            pitch, step, duration = predict_next_note(input_notes, model, temperature)
            start = prev_start + step
            end = start + duration
            input_note = (pitch, step, duration)
            generated_notes.append((*input_note, start, end))
            input_notes = np.delete(input_notes, 0, axis=0)
            input_notes = np.append(input_notes, np.expand_dims(input_note, 0), axis=0)
            prev_start = start
    
        generated_notes = pd.DataFrame(
            generated_notes, columns=(*key_order, 'start', 'end'))
    
        notes_to_midi(
            generated_notes, out_file=out_file, instrument_name=instrument_name)
    
        # Interfaz de Streamlit
        st.title("Generador de notas musicales")
    
        archivo_midi = open(out_file, 'rb').read()
        
        st.download_button(
            label="Descargar MIDI",
            data=archivo_midi,
            file_name=out_file,  # Nombre del archivo que se descargará
            mime='audio/midi'  
        )

        # Duracion del MIDI resultante
        duracion_f = calcular_duracion_midi(sample_file)
        minutos_f, segundos_f = divmod(duracion_f, 60)
        
        st.write(f"La duración del archivo MIDI es: {int(minutos_f)} minutos y {segundos_f:.2f} segundos.")

if __name__ == "__main__":
    main()