mapleafgo
4ddecb7c30
问题: - TTS 返回 16000 Hz 音频，但 Context 使用 44100 Hz - 播放速度快 2.75 倍（44100/16000） - 不同采样率的音频播放速度不正确 解决方案: - 集成 gomplerate 库（纯 Go，零依赖） - 自动检测音频采样率并重采样到 44100 Hz - 支持任意采样率的音频文件正常播放 技术实现: - resampler.go: 封装 gomplerate，实现流式重采样 - play.go: WAV/MP3 播放自动重采样 - loop.go: BGM 循环播放支持重采样 测试: - 所有单元测试通过（6/6） - 支持采样率自动转换（如 16000 Hz → 44100 Hz） 依赖: - github.com/zeozeozeo/gomplerate v0.0.0
143 lines
3.3 KiB
Go
143 lines
3.3 KiB
Go
package audio
|
|
|
|
import (
|
|
"bytes"
|
|
"context"
|
|
"fmt"
|
|
"io"
|
|
"time"
|
|
|
|
"github.com/youpy/go-wav"
|
|
"github.com/hajimehoshi/go-mp3"
|
|
"go.uber.org/zap"
|
|
)
|
|
|
|
// PlayWav 播放 WAV 文件(阻塞),直到完成或 context 取消
|
|
func PlayWav(ctx context.Context, r io.ReadCloser) error {
|
|
otoCtx, err := initContext()
|
|
if err != nil {
|
|
return fmt.Errorf("音频上下文初始化失败: %w", err)
|
|
}
|
|
|
|
// Read the entire file into memory since wav.NewReader needs ReadAt
|
|
data, err := io.ReadAll(r)
|
|
if err != nil {
|
|
r.Close()
|
|
return fmt.Errorf("读取 WAV 文件失败: %w", err)
|
|
}
|
|
r.Close()
|
|
|
|
// Create a reader from the buffered data
|
|
dec := wav.NewReader(bytes.NewReader(data))
|
|
|
|
// 获取音频格式信息
|
|
format, err := dec.Format()
|
|
if err != nil {
|
|
return fmt.Errorf("获取 WAV 格式失败: %w", err)
|
|
}
|
|
|
|
duration, _ := dec.Duration()
|
|
sourceRate := int(format.SampleRate)
|
|
targetRate := UniversalSampleRate
|
|
channels := int(format.NumChannels)
|
|
|
|
zap.S().Infof("WAV 音频: %d ch, %d Hz → %d Hz, 时长: %v",
|
|
channels, sourceRate, targetRate, duration)
|
|
|
|
// 需要重采样
|
|
var reader io.Reader = dec
|
|
if needsResample(sourceRate, targetRate) {
|
|
zap.S().Infof("重采样: %d Hz → %d Hz", sourceRate, targetRate)
|
|
resampleReader, err := newResamplingReader(dec, sourceRate, targetRate, channels)
|
|
if err != nil {
|
|
return fmt.Errorf("创建重采样器失败: %w", err)
|
|
}
|
|
reader = resampleReader
|
|
}
|
|
|
|
player := otoCtx.NewPlayer(reader)
|
|
defer player.Close()
|
|
|
|
player.Play()
|
|
|
|
// 等待播放完成
|
|
done := make(chan struct{})
|
|
go func() {
|
|
for !player.IsPlaying() {
|
|
time.Sleep(10 * time.Millisecond)
|
|
}
|
|
for player.IsPlaying() {
|
|
time.Sleep(10 * time.Millisecond)
|
|
}
|
|
time.Sleep(200 * time.Millisecond)
|
|
close(done)
|
|
}()
|
|
|
|
select {
|
|
case <-done:
|
|
return nil
|
|
case <-ctx.Done():
|
|
return ctx.Err()
|
|
}
|
|
}
|
|
|
|
// PlayMP3 播放 MP3 文件(阻塞),直到完成或 context 取消
|
|
func PlayMP3(ctx context.Context, r io.ReadCloser) error {
|
|
otoCtx, err := initContext()
|
|
if err != nil {
|
|
return fmt.Errorf("音频上下文初始化失败: %w", err)
|
|
}
|
|
|
|
dec, err := mp3.NewDecoder(r)
|
|
if err != nil {
|
|
r.Close()
|
|
return fmt.Errorf("MP3 解码失败: %w", err)
|
|
}
|
|
defer r.Close()
|
|
|
|
// MP3 解码器信息
|
|
sampleRate := int(dec.SampleRate())
|
|
sampleCount := dec.Length()
|
|
targetRate := UniversalSampleRate
|
|
channels := 2 // MP3 通常是立体声
|
|
duration := time.Duration(float64(sampleCount)/float64(sampleRate)*1000) * time.Millisecond
|
|
|
|
zap.S().Infof("MP3 音频: %d Hz → %d Hz, 时长约: %v", sampleRate, targetRate, duration)
|
|
|
|
// 需要重采样
|
|
var reader io.Reader = dec
|
|
if needsResample(sampleRate, targetRate) {
|
|
zap.S().Infof("重采样: %d Hz → %d Hz", sampleRate, targetRate)
|
|
resampleReader, err := newResamplingReader(dec, sampleRate, targetRate, channels)
|
|
if err != nil {
|
|
return fmt.Errorf("创建重采样器失败: %w", err)
|
|
}
|
|
reader = resampleReader
|
|
}
|
|
|
|
player := otoCtx.NewPlayer(reader)
|
|
defer player.Close()
|
|
|
|
player.Play()
|
|
|
|
// 等待播放完成
|
|
done := make(chan struct{})
|
|
go func() {
|
|
for !player.IsPlaying() {
|
|
time.Sleep(10 * time.Millisecond)
|
|
}
|
|
for player.IsPlaying() {
|
|
time.Sleep(10 * time.Millisecond)
|
|
}
|
|
time.Sleep(200 * time.Millisecond)
|
|
close(done)
|
|
}()
|
|
|
|
select {
|
|
case <-done:
|
|
return nil
|
|
case <-ctx.Done():
|
|
return ctx.Err()
|
|
}
|
|
}
|