game-driver/pkg/tts/aliyun.go

package tts

import (
	"bytes"
	"context"
	"fmt"
	"game-driver/config"
	"game-driver/leaf"
	"game-driver/pkg/audio"
	"game-driver/pkg/errorsx"
	"go.uber.org/zap"
	"io"
	"log"
	"time"

	nls "github.com/aliyun/alibabacloud-nls-go-sdk"
)

type AliTTS struct {
	config.AliyunConfig
	ctx         context.Context
	tokenResult nls.TokenResult
}

type result struct {
	Data  io.ReadWriter
	Error error
}

var DefaultTTS = &AliTTS{}

// onTaskFailed 识别过程中的错误处理回调参数
func (tts *AliTTS) onTaskFailed(text string, param interface{}) {
	p, _ := param.(*result)
	p.Error = fmt.Errorf("语音合成异常: %v", text)
}

// onSynthesisResult 语音合成数据回调参数
func (tts *AliTTS) onSynthesisResult(data []byte, param interface{}) {
	p, _ := param.(*result)
	p.Data.Write(data)
}

func (tts *AliTTS) Sound(text string) {
	if text == "" {
		return
	}
	zap.S().Infof("开始播放TTS: %s", text)
	buf, err := tts.Get(text)
	if err == nil && buf != nil {
		audio.PlayWav(tts.ctx, buf)
		zap.S().Infof("TTS播放完成: %s", text)
	} else {
		zap.S().Errorln("AliTTS 请求异常: ", err)
	}
}

func (tts *AliTTS) getToken() error {
	if tts.tokenResult.ExpireTime != 0 && time.Unix(tts.tokenResult.ExpireTime, 0).After(time.Now()) {
		return nil
	}
	tts.tokenResult = nls.TokenResult{}
	resultMessage, err := nls.GetToken("cn-shanghai", "nls-meta.cn-shanghai.aliyuncs.com", tts.AccessKeyID, tts.AccessKeySecret, "2019-02-28")
	if err != nil {
		return err
	} else if resultMessage.ErrMsg != "" {
		zap.S().Errorf("获取Token失败: %s", resultMessage.ErrMsg)
		return errorsx.ThirdPartyErr
	}
	tts.tokenResult = resultMessage.TokenResult
	return nil
}

func (tts *AliTTS) Get(text string) (io.Reader, error) {
	param := nls.DefaultSpeechSynthesisParam()
	param.Volume = tts.Volume
	param.Voice = tts.Voice
	param.SpeechRate = tts.SpeechRate

	err := tts.getToken()
	if err != nil {
		return nil, err
	}

	connectConfig := nls.NewConnectionConfigWithToken(nls.DEFAULT_URL, tts.AppKey, tts.tokenResult.Id)

	logger := nls.NewNlsLogger(leaf.DefaultWriter, "", log.LstdFlags|log.Ltime)
	logger.SetLogSil(false)
	logger.SetDebug(true)

	ttsData := &result{
		Data: &bytes.Buffer{},
	}

	synthesis, err := nls.NewSpeechSynthesis(
		connectConfig, logger, false,
		tts.onTaskFailed, tts.onSynthesisResult, nil,
		nil, nil, ttsData,
	)
	if err != nil {
		return ttsData.Data, err
	}
	defer synthesis.Shutdown()

	ch, err := synthesis.Start(text, param, nil)
	if err != nil {
		return ttsData.Data, err
	}

	// 等待语音合成结束
	select {
	case done := <-ch:
		{
			if !done {
				return ttsData.Data, errorsx.ThirdPartyErr
			}
			return ttsData.Data, nil
		}
	case <-time.After(time.Duration(tts.Timeout) * time.Second):
		return ttsData.Data, errorsx.DriverTimeoutErr
	case <-tts.ctx.Done():
		return ttsData.Data, errorsx.DriverCancelErr
	}
}

func New(ctx context.Context, config config.AliyunConfig) *AliTTS {
	return &AliTTS{
		ctx:          ctx,
		AliyunConfig: config,
	}
}