简介

AI音乐生成利用深度学习模型从文本描述或旋律片段创作音乐。本节介绍MusicGen、Suno AI等工具的使用方法和音乐Prompt设计技巧。

MusicGen使用

# 使用Meta MusicGen生成音乐
from audiocraft.models import MusicGen
from audiocraft.data.audio import audio_write
import torch

# 加载预训练模型
model = MusicGen.get_pretrained("medium")
model.set_generation_params(duration=30)  # 生成30秒

def generate_music(descriptions, output_dir="music_output"):
    """根据文本描述生成音乐"""
    import os
    os.makedirs(output_dir, exist_ok=True)
    
    wav = model.generate(descriptions)
    for idx, (desc, one_wav) in enumerate(zip(descriptions, wav)):
        filename = f"music_{idx}"
        audio_write(
            f"{output_dir}/{filename}",
            one_wav.cpu(),
            model.sample_rate,
            strategy="loudness"
        )
        print(f"已生成: {output_dir}/{filename}.wav (Prompt: {desc[:50]}...)")

descriptions = [
    "upbeat electronic dance music with strong bass, 128 BPM, festival style",
    "calm piano melody for meditation, soft and peaceful, ambient",
    "epic orchestral battle music, dramatic strings and drums"
]

generate_music(descriptions)

音乐Prompt设计

维度	关键词	效果
风格	electronic, classical, jazz, rock, lo-fi	决定音乐类型
情绪	upbeat, calm, dramatic, melancholic, happy	控制情感基调
乐器	piano, guitar, synthesizer, drums, violin	指定主要乐器
节奏	128 BPM, slow tempo, fast-paced	控制速度感
场景	cinematic, festival, coffee shop, meditation	营造空间感

Suno AI API调用

# Suno AI API调用示例（非官方SDK）
import requests

SUNO_API = "https://api.suno.ai/v1/generation"
API_KEY = "your-api-key"

def generate_song(prompt, tags="", title=""):
    response = requests.post(SUNO_API, headers={
        "Authorization": f"Bearer {API_KEY}",
        "Content-Type": "application/json"
    }, json={
        "prompt": prompt,
        "tags": tags,        # 音乐风格标签
        "title": title,
        "make_instrumental": False,
        "wait_audio": True    # 等待生成完成
    })
    return response.json()

result = generate_song(
    prompt="一首关于编程的流行歌曲，歌词包含代码和咖啡的意象",
    tags="pop, upbeat",
    title="Code and Coffee"
)
print(f"歌曲URL: {result[0]["audio_url"]}")

音频后处理

# 使用pydub处理生成的音频
from pydub import AudioSegment

def process_audio(input_path, output_path):
    audio = AudioSegment.from_wav(input_path)
    # 标准化音量
    audio = audio.normalize()
    # 淡入淡出
    audio = audio.fade_in(2000).fade_out(3000)
    # 导出为MP3
    audio.export(output_path, format="mp3", bitrate="192k")
    print(f"已导出: {output_path}")

process_audio("music_output/music_0.wav", "final_music.mp3")

小结

AI音乐生成的关键是Prompt中对风格、情绪、乐器和节奏的精确描述。MusicGen适合生成背景音乐和配乐，Suno AI能生成含人声的完整歌曲。注意版权问题，AI生成的音乐可能受使用条款限制。

移动AI编程入门