CorpusAI/service/text_to_speech.go

package service

import (
	"bytes"
	"cloud.google.com/go/texttospeech/apiv1"
	"crypto/md5"
	"fmt"
	"git.links123.net/Slate/CorpusAI/config"
	"github.com/aliyun/aliyun-oss-go-sdk/oss"
	texttospeechpb "google.golang.org/genproto/googleapis/cloud/texttospeech/v1"
	"golang.org/x/net/context"
)

// @Description Get TTS Oss unique key
func GetTtsOssKey(Text string, VoiceName string, LanguageCode string, Speed float64, Pitch float64) string {

	data := []byte(Text)
	hash := md5.Sum(data)
	md5Value := fmt.Sprintf("%x", hash)

	key := fmt.Sprintf("%s/%s/%s/%s/%f_%f.mp3", "tts/en/google", md5Value, LanguageCode, VoiceName, Speed, Pitch)

	return key
}

func GetUrl(ossObjectKey string) string {

	ossConfig := config.C.OSS

	appConfig := config.C.App
	if appConfig.InChina == true {

		return fmt.Sprintf("https://%s.%s/%s", ossConfig.Bucket, ossConfig.EndPoint, ossObjectKey)

	} else {
		return fmt.Sprintf("https://%s.%s/%s", ossConfig.HkBucket, ossConfig.HkEndPoint, ossObjectKey)
	}
}

// @Description Upload AudioContent to Oss
func UploadOss(ossObjectKey string, AudioContent []byte) (bool , error) {
	ossConfig := config.C.OSS
	client, err := oss.New(ossConfig.EndPoint, ossConfig.AccessKey, ossConfig.SecretKey)
	bucket, err := client.Bucket(ossConfig.Bucket)
	if err != nil {
		return false, fmt.Errorf("texttospeech.UploadOss Bucket error: %v", err)
	}

	// 上传Byte数组。
	err = bucket.PutObject(ossObjectKey, bytes.NewReader(AudioContent))
	isExist, err := bucket.IsObjectExist(ossObjectKey)
	if err != nil {
		return false, fmt.Errorf("texttospeech.UploadOss Put error: %v", err)
	}
	return isExist, nil
}

func UploadHkOss(ossObjectKey string, AudioContent []byte) (bool , error) {
	ossConfig := config.C.OSS
	client, err := oss.New(ossConfig.HkEndPoint, ossConfig.AccessKey, ossConfig.SecretKey)
	bucket, err := client.Bucket(ossConfig.HkBucket)
	if err != nil {
		return false, fmt.Errorf("texttospeech.UploadOss Bucket error: %v", err)
	}

	// 上传Byte数组。
	err = bucket.PutObject(ossObjectKey, bytes.NewReader(AudioContent))
	isExist, err := bucket.IsObjectExist(ossObjectKey)
	if err != nil {
		return false, fmt.Errorf("texttospeech.UploadOss Put error: %v", err)
	}
	return isExist, nil
}

// @Description Text to speech
func TextToSpeech(Text string, VoiceName string, LanguageCode string, Speed float64, Pitch float64) ([]byte , error) {

	ctx := context.Background()

	client, err := texttospeech.NewClient(ctx)
	if err != nil {
		return nil, fmt.Errorf("texttospeech.NewClient error: %v", err)
	}

	req := texttospeechpb.SynthesizeSpeechRequest{
		// Set the text input to be synthesized.
		Input: &texttospeechpb.SynthesisInput{
			InputSource: &texttospeechpb.SynthesisInput_Text{Text: Text},
		},
		// Build the voice request, select the language code ("en-US") and the SSML
		// voice gender ("neutral").
		Voice: &texttospeechpb.VoiceSelectionParams{
			Name: VoiceName,
			LanguageCode: LanguageCode,
			//SsmlGender:   texttospeechpb.SsmlVoiceGender_NEUTRAL,
		},
		// Select the type of audio file you want returned.
		AudioConfig: &texttospeechpb.AudioConfig{
			AudioEncoding: texttospeechpb.AudioEncoding_MP3,
			SpeakingRate: Speed,
			Pitch: Pitch,
		},
	}

	resp, err := client.SynthesizeSpeech(ctx, &req)
	if err != nil {
		return nil, fmt.Errorf("texttospeech.SynthesizeSpeech error: %v", err)
	}

	return resp.AudioContent, nil
}