AI写歌源代码 - 人工智能音乐创作编程指南

# 安装必要的依赖包 pip install tensorflow keras mido music21 numpy matplotlib # 导入核心库 import tensorflow as tf from tensorflow import keras import numpy as np import mido from music21 import converter, instrument, note, chord, stream import random

def prepare_sequences(notes, pitchnames, n_vocab): """ 准备训练序列数据 """ # 创建音符到整数的映射 note_to_int = dict((note, number) for number, note in enumerate(pitchnames)) sequence_length = 100 network_input = [] network_output = [] # 创建输入序列和目标输出 for i in range(0, len(notes) - sequence_length, 1): sequence_in = notes[i:i + sequence_length] sequence_out = notes[i + sequence_length] network_input.append([note_to_int[char] for char in sequence_in]) network_output.append(note_to_int[sequence_out]) n_patterns = len(network_input) # 重塑输入为LSTM格式 [样本数, 时间步长, 特征数] normalized_input = np.reshape(network_input, (n_patterns, sequence_length, 1)) # 归一化 normalized_input = normalized_input / float(n_vocab) return (network_input, normalized_input) def create_midi(prediction_output): """ 将预测结果转换为MIDI文件 """ offset = 0 output_notes = [] # 创建音符和和弦对象列表 for pattern in prediction_output: # 如果是和弦 if ('.' in pattern) or pattern.isdigit(): notes_in_chord = pattern.split('.') notes = [] for current_note in notes_in_chord: new_note = note.Note(int(current_note)) new_note.storedInstrument = instrument.Piano() notes.append(new_note) new_chord = chord.Chord(notes) new_chord.offset = offset output_notes.append(new_chord) # 如果是单音符 else: new_note = note.Note(pattern) new_note.offset = offset new_note.storedInstrument = instrument.Piano() output_notes.append(new_note) offset += 0.5 # 创建音乐流 midi_stream = stream.Stream(output_notes) return midi_stream

def create_network(network_input, n_vocab): """ 创建LSTM神经网络模型 """ model = keras.Sequential([ keras.layers.LSTM( 512, input_shape=(network_input.shape[1], network_input.shape[2]), recurrent_dropout=0.3, return_sequences=True ), keras.layers.Dropout(0.3), keras.layers.LSTM(512, return_sequences=True, recurrent_dropout=0.3), keras.layers.Dropout(0.3), keras.layers.LSTM(512), keras.layers.Dense(256), keras.layers.Dropout(0.3), keras.layers.Dense(n_vocab), keras.layers.Activation('softmax') ]) model.compile(loss='categorical_crossentropy', optimizer='rmsprop') return model def train_network(): """ 训练神经网络模型 """ # 加载和处理训练数据 notes = get_notes() # 自定义函数：从MIDI文件提取音符 pitchnames = sorted(set(item for item in notes)) n_vocab = len(set(notes)) network_input, normalized_input = prepare_sequences(notes, pitchnames, n_vocab) model = create_network(normalized_input, n_vocab) # 训练模型 filepath = "weights-improvement-{epoch:02d}-{loss:.4f}-bigger.hdf5" checkpoint = keras.callbacks.ModelCheckpoint( filepath, monitor='loss', verbose=0, save_best_only=True, mode='min' ) callbacks_list = [checkpoint] model.fit(network_input, network_output, epochs=200, batch_size=64, callbacks=callbacks_list)

def style_transfer_composition(base_model, style_models, weights=[0.5, 0.3, 0.2]): """ 多风格融合作曲 """ compositions = [] for model, weight in zip(style_models, weights): comp = generate_from_model(model) compositions.append(comp * weight) # 融合不同风格的作品 final_composition = blend_compositions(compositions) return final_composition