feat:Add pitch Option

Author: czyt

internal/communicate/communicate.go

@@ -28,7 +28,7 @@ import (
 
 const (
 	ssmlHeaderTemplate = "X-RequestId:%s\r\nContent-Type:application/ssml+xml\r\nX-Timestamp:%sZ\r\nPath:ssml\r\n\r\n"
-	ssmlTemplate       = "<speak version='1.0' xmlns='http://www.w3.org/2001/10/synthesis' xml:lang='en-US'><voice name='%s'><prosody pitch='+0Hz' rate='%s' volume='%s'>%s</prosody></voice></speak>"
+	ssmlTemplate       = "<speak version='1.0' xmlns='http://www.w3.org/2001/10/synthesis' xml:lang='en-US'><voice name='%s'><prosody pitch='%s' rate='%s' volume='%s'>%s</prosody></voice></speak>"
 )
 
 var (
@@ -46,26 +46,27 @@ func init() {
 type Communicate struct {
 	text                string
 	voice               string
-	voiceLanguageRegion string
+	pitch               string
 	rate                string
 	volume              string
+	voiceLanguageRegion string
 
 	httpProxy        string
 	socket5Proxy     string
 	socket5ProxyUser string
 	socket5ProxyPass string
-	op               chan map[string]interface{}
 
 	audioDataIndex int
 	prevIdx        int
 	shiftTime      int
 	finalUtterance map[int]int
+	op             chan map[string]interface{}
 }
 
 type textEntry struct {
 	Text         string `json:"text"`
-	Length       int64  `json:"Length"`
 	BoundaryType string `json:"BoundaryType"`
+	Length       int64  `json:"Length"`
 }
 type dataEntry struct {
 	Offset   int       `json:"Offset"`
@@ -112,6 +113,7 @@ func NewCommunicate(text string, opt *communicateOption.CommunicateOption) (*Com
 	}
 	return &Communicate{
 		text:                text,
+		pitch:               opt.Pitch,
 		voice:               opt.Voice,
 		voiceLanguageRegion: opt.VoiceLangRegion,
 		rate:                opt.Rate,
@@ -171,7 +173,7 @@ func makeHeaders() http.Header {
 func (c *Communicate) stream() (<-chan map[string]interface{}, error) {
 	texts := splitTextByByteLength(
 		escape(removeIncompatibleCharacters(c.text)),
-		calculateMaxMessageSize(c.voice, c.rate, c.volume),
+		calculateMaxMessageSize(c.pitch, c.voice, c.rate, c.volume),
 	)
 	c.audioDataIndex = len(texts)
 
@@ -234,7 +236,7 @@ func (c *Communicate) sendSSML(conn *websocket.Conn, currentTime string, text []
 			ssmlHeadersAppendExtraData(
 				generateConnectID(),
 				currentTime,
-				makeSsml(string(text), c.voice, c.rate, c.volume),
+				makeSsml(string(text), c.pitch, c.voice, c.rate, c.volume),
 			),
 		))
 }
@@ -474,10 +476,11 @@ func splitTextByByteLength(text string, byteLength int) [][]byte {
 	return result
 }
 
-func makeSsml(text string, voice string, rate string, volume string) string {
+func makeSsml(text string, pitch, voice string, rate string, volume string) string {
 	ssml := fmt.Sprintf(
 		ssmlTemplate,
 		voice,
+		pitch,
 		rate,
 		volume,
 		text)
@@ -500,9 +503,9 @@ func ssmlHeadersAppendExtraData(requestID string, timestamp string, ssml string)
 	return headers + ssml
 }
 
-func calculateMaxMessageSize(voice string, rate string, volume string) int {
+func calculateMaxMessageSize(pitch, voice string, rate string, volume string) int {
 	websocketMaxSize := 1 << 16
-	overheadPerMessage := len(ssmlHeadersAppendExtraData(generateConnectID(), currentTimeInMST(), makeSsml("", voice, rate, volume))) + 50
+	overheadPerMessage := len(ssmlHeadersAppendExtraData(generateConnectID(), currentTimeInMST(), makeSsml("", pitch, voice, rate, volume))) + 50
 	return websocketMaxSize - overheadPerMessage
 }
 

internal/communicateOption/option.go

@@ -2,13 +2,15 @@ package communicateOption
 
 import (
 	"fmt"
-	"github.com/lib-x/edgetts/internal/businessConsts"
 	"strings"
+
+	"github.com/lib-x/edgetts/internal/businessConsts"
 )
 
 type CommunicateOption struct {
 	Voice            string
 	VoiceLangRegion  string
+	Pitch            string
 	Rate             string
 	Volume           string
 	HttpProxy        string
@@ -31,6 +33,9 @@ func (c *CommunicateOption) CheckAndApplyDefaultOption() {
 		name := voiceParsed[2]
 		c.VoiceLangRegion = fmt.Sprintf(businessConsts.VoiceNameTemplate, lang, region, name)
 	}
+	if c.Pitch == "" {
+		c.Pitch = "+0Hz"
+	}
 	if c.Rate == "" {
 		c.Rate = "+0%"
 	}

internal/validate/validate.go

@@ -2,17 +2,20 @@ package validate
 
 import (
 	"errors"
-	"github.com/lib-x/edgetts/internal/communicateOption"
 	"regexp"
+
+	"github.com/lib-x/edgetts/internal/communicateOption"
 )
 
 var (
+	validPitchPattern      = regexp.MustCompile(`^[+-]\d+Hz$`)
 	validVoicePattern      = regexp.MustCompile(`^([a-z]{2,})-([A-Z]{2,})-(.+Neural)$`)
 	validRateVolumePattern = regexp.MustCompile(`^[+-]\d+%$`)
 )
 
 var (
 	InvalidVoiceError  = errors.New("invalid voice")
+	InvalidPitchError  = errors.New("invalid pitch")
 	InvalidRateError   = errors.New("invalid rate")
 	InvalidVolumeError = errors.New("invalid volume")
 )
@@ -24,8 +27,11 @@ func WithCommunicateOption(c *communicateOption.CommunicateOption) error {
 		return InvalidVoiceError
 	}
 
+	// WithCommunicateOption pitch
+	if !validPitchPattern.MatchString(c.Pitch) {
+		return InvalidPitchError
+	}
 	// WithCommunicateOption rate
-
 	if !validRateVolumePattern.MatchString(c.Rate) {
 		return InvalidRateError
 	}
@@ -34,5 +40,6 @@ func WithCommunicateOption(c *communicateOption.CommunicateOption) error {
 	if !validRateVolumePattern.MatchString(c.Volume) {
 		return InvalidVolumeError
 	}
+
 	return nil
 }

option.go

@@ -5,6 +5,7 @@ import "github.com/lib-x/edgetts/internal/communicateOption"
 type option struct {
 	Voice            string
 	VoiceLangRegion  string
+	Pitch            string
 	Rate             string
 	Volume           string
 	HttpProxy        string
@@ -17,6 +18,7 @@ func (o *option) toInternalOption() *communicateOption.CommunicateOption {
 	return &communicateOption.CommunicateOption{
 		Voice:            o.Voice,
 		VoiceLangRegion:  o.VoiceLangRegion,
+		Pitch:            o.Pitch,
 		Rate:             o.Rate,
 		Volume:           o.Volume,
 		HttpProxy:        o.HttpProxy,
@@ -41,12 +43,21 @@ func WithVoiceLangRegion(voiceLangRegion string) Option {
 
 }
 
+// WithPitch set pitch of the tts output.such as +50Hz,-50Hz
+func WithPitch(pitch string) Option {
+	return func(option *option) {
+		option.Pitch = pitch
+	}
+}
+
+// WithRate set rate of the tts output.rate=-50% means rate down 50%,rate=+50% means rate up 50%
 func WithRate(rate string) Option {
 	return func(option *option) {
 		option.Rate = rate
 	}
 }
 
+// WithVolume set volume of the tts output.volume=-50% means volume down 50%,volume=+50% means volume up 50%
 func WithVolume(volume string) Option {
 	return func(option *option) {
 		option.Volume = volume