Python使用pydub处理音频切片与格式转换的完整指南

pip install pydub

poetry add pydub
# 或
uv add pydub

import pydub
print(pydub.__version__)  # 输出类似 0.25.1

from importlib.metadata import version
print(version("pydub"))

winget install ffmpeg

ffmpeg -version

brew install ffmpeg

sudo apt update && sudo apt install ffmpeg

sudo yum install epel-release
sudo yum install ffmpeg

from pydub import audiosegment
from pydub.utils import which

# 方式一：直接设置可执行文件路径
audiosegment.converter = "/usr/local/bin/ffmpeg"
audiosegment.ffprobe = "/usr/local/bin/ffprobe"

# 方式二：自动探测（推荐）
print(which("ffmpeg"))  # 返回 ffmpeg 路径，若为 none 则说明未找到

from pydub import audiosegment
from pydub.playback import play  # 需要额外安装 simpleaudio

# 生成一段 1 秒的静音音频（44100hz，双声道）
silence = audiosegment.silent(duration=1000)
print(f"时长: {len(silence)}ms, 声道数: {silence.channels}, 帧率: {silence.frame_rate}hz")

# 导出为 wav 测试 ffmpeg 是否正常工作
silence.export("test.wav", format="wav")
print("ffmpeg 工作正常")

pip install pydub

from pydub import audiosegment

song = audiosegment.from_file("demo.mp3", format="mp3")
wav_audio = audiosegment.from_wav("demo.wav")          # 等价于 format="wav"
raw_ogg = audiosegment.from_ogg("demo.ogg")            # 等价于 format="ogg"

import wave

with wave.open("voice.wav", "rb") as wf:
    raw_data = wf.readframes(wf.getnframes())
    seg = audiosegment(
        data=raw_data,
        sample_width=wf.getsampwidth(),   # 2 表示 16-bit
        frame_rate=wf.getframerate(),     # 通常为 44100 或 48000
        channels=wf.getnchannels()        # 1 为单声道，2 为立体声
    )

print(seg.frame_rate)        # 44100，每秒采样帧数
print(seg.channels)          # 2，声道数
print(seg.sample_width)      # 2，每个采样点的字节数（1=8bit，2=16bit，4=32bit float）
print(seg.duration_seconds)  # 3.5，音频总时长（秒，浮点数）

# 由上述属性推导出的常用量
total_frames = seg.frame_count()   # 总帧数 = duration_seconds * frame_rate
byte_rate = seg.frame_rate * seg.channels * seg.sample_width  # 每秒字节数

seg_16bit = seg.set_sample_width(2)   # 返回新对象，原对象不变

from pydub import audiosegment

# 加载音频文件，pydub 会根据扩展名自动选择解码器
song = audiosegment.from_file("input.mp3", format="mp3")
# 也可以直接用专用方法，效果相同
song = audiosegment.from_mp3("input.mp3")

# 截取第 10 秒到第 20 秒（10000ms ~ 20000ms）
clip = song[10000:20000]

# 省略 start，表示从开头到第 5 秒
head = song[:5000]

# 省略 end，表示从第 30 秒到结尾
tail = song[30000:]

# 支持负索引，从结尾倒数
last_3s = song[-3000:]

intro = audiosegment.from_wav("intro.wav")
outro = audiosegment.from_wav("outro.wav")

# 拼接后生成完整音频
full = intro + outro

# 也可以连续拼接
combined = song[:5000] + song[10000:15000] + song[-3000:]

mono = audiosegment.from_wav("mono.wav")
stereo = mono.set_channels(2)  # 转为双声道
ok = stereo + outro  # 现在可以正常拼接

# 导出为 wav（无损）
clip.export("output.wav", format="wav")

# 导出为 mp3，可指定比特率
clip.export("output.mp3", format="mp3", bitrate="192k")

# 导出为 ogg 格式
clip.export("output.ogg", format="ogg")

# 导出到内存中的文件对象
from io import bytesio
buf = bytesio()
clip.export(buf, format="mp3")
buf.seek(0)  # 方便后续上传或网络传输

from pydub import audiosegment

source = audiosegment.from_file("podcast.mp3")

# 切出三个片段：开头 15 秒、中间 30 秒、结尾 10 秒
segments = {
    "opening": source[:15000],
    "middle": source[60000:90000],
    "ending": source[-10000:],
}

for name, seg in segments.items():
    # 统一导出为 wav，便于后续编辑
    seg.export(f"{name}.wav", format="wav")
    # 同时生成 ogg 用于网页播放
    seg.export(f"{name}.ogg", format="ogg")

pip install pydub

from pydub import audiosegment
from pathlib import path
import sys

def cut_and_convert(input_path: str, output_path: str, duration_ms: int = 30_000):
    """
    从 mp3 截取前 duration_ms 毫秒并导出为 wav。
    
    参数:
        input_path: 源 mp3 文件路径
        output_path: 输出 wav 文件路径
        duration_ms: 截取时长，默认 30000ms（30秒）
    
    返回:
        bool: 成功返回 true，失败返回 false
    """
    try:
        # 1. 检查源文件是否存在
        if not path(input_path).exists():
            raise filenotfounderror(f"输入文件不存在: {input_path}")
        
        # 2. 读取 mp3 —— audiosegment 会自动调用 ffmpeg
        audio = audiosegment.from_mp3(input_path)
        print(f"原始音频时长: {len(audio)} ms, 声道数: {audio.channels}, 采样率: {audio.frame_rate} hz")
        
        # 3. 切片：取前 duration_ms 毫秒
        #    如果原音频不足 30 秒，取 min 避免越界
        actual_duration = min(duration_ms, len(audio))
        segment = audio[:actual_duration]
        
        # 4. 导出 wav
        #    参数 format='wav' 指定容器，bitrate 对 wav 不适用，但可设 sample_width
        segment.export(output_path, format="wav")
        print(f"成功导出: {output_path} (实际时长 {len(segment)} ms)")
        return true
        
    except filenotfounderror as e:
        print(f"文件错误: {e}", file=sys.stderr)
    except exception as e:
        # 捕获 ffmpeg 缺失、解码失败等一切异常
        print(f"处理失败: {type(e).__name__}: {e}", file=sys.stderr)
        print("若提示 ffmpeg 相关错误，请先安装 ffmpeg 并确认可被系统找到。", file=sys.stderr)
    return false


if __name__ == "__main__":
    # 用法示例
    success = cut_and_convert("podcast.mp3", "intro_30s.wav")
    if not success:
        sys.exit(1)

原始音频时长: 200000 ms, 声道数: 2, 采样率: 44100 hz
成功导出: intro_30s.wav (实际时长 30000 ms)

from pydub import audiosegment

song = audiosegment.from_file("intro.mp3", format="mp3")

# 音量提升 3db，适合响度偏低的背景音乐
louder = song.apply_gain(3.0)
# 音量衰减 10db，适合做低音量垫底
quieter = song.apply_gain(-10.0)

# 导出验证
louder.export("intro_louder.mp3", format="mp3")
quieter.export("intro_quieter.mp3", format="mp3")

# 从第 10 秒开始截取 8 秒，并做 1 秒淡入、2 秒淡出
clip = song[10000:18000].fade_in(1000).fade_out(2000)
clip.export("clip_with_fade.mp3", format="mp3")

# 准备三段短音频
part1 = audiosegment.from_file("part1.mp3")
part2 = audiosegment.from_file("part2.mp3")
part3 = audiosegment.from_file("part3.mp3")

# 统一为单声道、44.1khz、16bit，避免拼接异常
part1 = part1.set_channels(1).set_frame_rate(44100).set_sample_width(2)
part2 = part2.set_channels(1).set_frame_rate(44100).set_sample_width(2)
part3 = part3.set_channels(1).set_frame_rate(44100).set_sample_width(2)

combined = part1 + part2 + part3
combined.export("combined.mp3", format="mp3")

from pydub import audiosegment

source = audiosegment.from_file("source.wav")

# 提取两个片段：0-3秒和5-8秒
melody_a = source[0:3000].fade_in(200).fade_out(300)
melody_b = source[5000:8000].fade_in(150).fade_out(250)

# 统一参数后拼接，再整体循环两次
melody_a = melody_a.set_channels(1).set_frame_rate(44100)
melody_b = melody_b.set_channels(1).set_frame_rate(44100)

loop_sound = (melody_a + melody_b) * 2
loop_sound.export("loop_notification.wav", format="wav")

# ubuntu/debian
sudo apt-get install ffmpeg
# macos (homebrew)
brew install ffmpeg
# windows (choco)
choco install ffmpeg

from pydub import audiosegment
from pydub.utils import which

# 手动指定 ffmpeg 可执行文件路径
audiosegment.converter = "/usr/local/bin/ffmpeg"
audiosegment.ffprobe = "/usr/local/bin/ffprobe"

audio = audiosegment.from_file("input.mp3")
print(f"时长: {len(audio)}ms")  # 能走到这里说明 ffmpeg 配置成功

from pydub import audiosegment

audio = audiosegment.from_file("input.mp3")  # 假设时长 10000ms

# 安全切片：先 clamp 再切
def safe_slice(seg, start_ms, end_ms):
    start_ms = max(0, min(start_ms, len(seg)))
    end_ms = max(start_ms, min(end_ms, len(seg)))
    return seg[start_ms:end_ms]

part1 = safe_slice(audio, 5000, 999999)  # 不会报错，返回 5000ms 到结尾
part2 = safe_slice(audio, -3000, 2000)   # 负数被 clamp 到 0
print(f"part1: {len(part1)}ms, part2: {len(part2)}ms")

# 导出为浏览器兼容的 mp3 (固定码率 128kbps)
audio.export("out.mp3", format="mp3", bitrate="128k")

# 导出为 16bit pcm wav（默认已如此，但显式写出更清晰）
audio.export("out.wav", format="wav", parameters=["-acodec", "pcm_s16le"])

# 导出为 m4a (aac 编码) — 注意 ffmpeg 需要编译了 aac 支持
audio.export("out.m4a", format="ipod", codec="aac")

from pydub import audiosegment

def process_large_file(path, chunk_ms=30000):
    """按 30 秒分块处理，避免一次性加载全部内存"""
    audio = audiosegment.from_file(path)
    total_ms = len(audio)
    results = []
    
    for start in range(0, total_ms, chunk_ms):
        end = min(start + chunk_ms, total_ms)
        chunk = audio[start:end]
        # 对 chunk 做处理（例如降噪、增益）
        processed = chunk.apply_gain(-3)  # 示例操作
        results.append(processed)
    
    # 最后拼接
    return sum(results[1:], results[0]) if results else audiosegment.empty()

processed = process_large_file("long_audio.mp3")
processed.export("processed.mp3", format="mp3")

a = audiosegment.from_file("a.mp3").set_frame_rate(44100).set_channels(2)
b = audiosegment.from_file("b.wav").set_frame_rate(44100).set_channels(2)
combined = a + b  # 现在安全，不会触发隐式重采样

from pydub import audiosegment
from pydub.exceptions import couldntdecodeerror

try:
    audio = audiosegment.from_file("broken.mp3")
except couldntdecodeerror as e:
    print(f"解码失败，ffmpeg 报错: {e}")
    # 这里 e 通常包含 ffmpeg 的完整 stderr 输出，是定位问题的关键
except filenotfounderror:
    print("请检查 ffmpeg 是否安装并加入 path")

from pydub import audiosegment

song = audiosegment.from_file("interview.wav", format="wav")
# 取出第 10 秒到第 20 秒作为开头片段
intro = song[10_000:20_000]
# 取出最后 5 秒作为结尾
outro = song[-5_000:]
# 拼接并导出
result = intro + outro
result.export("teaser.mp3", format="mp3", bitrate="128k")

from pydub import audiosegment
from pydub.silence import split_on_silence

audio = audiosegment.from_file("call_center.mp3", format="mp3")
# 统一为 16khz 单声道，适配 asr 模型
audio = audio.set_frame_rate(16000).set_channels(1)
# 按静音切分（-35dbfs 以下视为静音，最短 500ms，前后保留 200ms）
chunks = split_on_silence(audio, min_silence_len=500, silence_thresh=-35, keep_silence=200)

for i, chunk in enumerate(chunks):
    chunk.export(f"segment_{i:03d}.wav", format="wav")

from pathlib import path
from pydub import audiosegment

input_dir = path("raw_audio")
output_dir = path("converted")
output_dir.mkdir(exist_ok=true)

for src in input_dir.glob("*.wav"):
    try:
        audio = audiosegment.from_file(src)
        # 转成 44.1khz 立体声 mp3
        audio.set_frame_rate(44100).set_channels(2).export(
            output_dir / f"{src.stem}.mp3", format="mp3"
        )
    except exception as e:
        print(f"跳过 {src.name}: {e}")

import numpy as np
from pydub import audiosegment

# 假设 y 是 librosa 处理后的 float32 数组，sr=22050
y = np.zeros(22050, dtype=np.float32)  # 1秒静音占位
# 转成 int16 并封装为 audiosegment
samples = (y * 32767).astype(np.int16)
audio = audiosegment(
    samples.tobytes(),
    frame_rate=22050,
    sample_width=2,  # int16 = 2字节
    channels=1
)
audio.export("out.wav", format="wav")