Initial support of aivisspeech

config-template.yaml

@@ -37,6 +37,9 @@ voices:
   coeiroink:
     enabled: false
     api:
+  aivisspeech:
+    enabled: false
+    api:
   aquestalkproxy:
     enabled: false
     api:

lib/config/config.go

@@ -55,6 +55,10 @@ type Config struct {
 			Enabled bool
 			Api     string
 		}
+		AivisSpeech struct {
+			Enabled bool
+			Api     string
+		}
 		AquestalkProxy struct {
 			Enabled bool
 			Api     string
@@ -100,7 +104,7 @@ func init() {
 		log.Fatal("Config parse failed:", err)
 	}
 
-	//verify
+	// verify
 	if CurrentConfig.Debug {
 		log.Print("Debug is enabled")
 	}
@@ -120,7 +124,7 @@ func init() {
 func VerifyGuild(guild *Guild) error {
 	val, exists := Lang[guild.Lang]
 	if !exists {
-		return errors.New("no such language") //Don't use nil val!
+		return errors.New("no such language") // Don't use nil val!
 	}
 	guilderrorstr := val.Error.Guild
 	if len(guild.Prefix) != 1 {

lib/voices/aivisspeech.go

@@ -0,0 +1,133 @@
+package voices
+
+import (
+	"bytes"
+	"encoding/json"
+	"errors"
+	"io"
+	"log"
+	"net/http"
+	"net/url"
+	"strconv"
+
+	"github.com/tpc3/Kotone-DiVE/lib/config"
+)
+
+var AivisSpeech aivisspeech
+
+type aivisspeech struct {
+	Info     VoiceInfo
+	Speakers Speakers
+	Request  *http.Request
+}
+
+// These structs are currently used from voicevox implementation
+// type Speakers struct
+// type Speaker struct
+// type Style struct
+
+// AivisSpeech is the another engine implementation based on the voicevox.
+// They uses almost identical api, but does not provide official docker image (There's Dockerfile though).
+
+func init() {
+	AivisSpeech = aivisspeech{
+		Info: VoiceInfo{
+			Type:             "aivisspeech",
+			Format:           "pcm",
+			Container:        "wav",
+			ReEncodeRequired: true,
+			Enabled:          config.CurrentConfig.Voices.AivisSpeech.Enabled,
+		},
+	}
+	if !config.CurrentConfig.Voices.AivisSpeech.Enabled {
+		log.Print("WARN: AivisSpeech is disabled")
+		return
+	}
+	res, err := http.Get(config.CurrentConfig.Voices.AivisSpeech.Api + "/speakers")
+	if err != nil {
+		log.Fatal(err)
+	}
+	defer res.Body.Close()
+	bin, err := io.ReadAll(res.Body)
+	if err != nil {
+		log.Fatal(err)
+	}
+	err = json.Unmarshal([]byte("{ \"speakers\": "+string(bin)+" }"), &AivisSpeech.Speakers)
+	if err != nil {
+		log.Fatal(err)
+	}
+	request, err := http.NewRequest(http.MethodPost, config.CurrentConfig.Voices.AivisSpeech.Api+"/synthesis", nil)
+	if err != nil {
+		log.Fatal(err)
+	}
+	AivisSpeech.Request = request
+}
+
+func (voiceSource aivisspeech) Synth(content string, voice string) ([]byte, error) {
+	id := -1
+	for _, speaker := range voiceSource.Speakers.Speakers {
+		for _, v := range speaker.Styles {
+			if speaker.Name+v.Name == voice {
+				id = v.Id
+				break
+			}
+		}
+	}
+	if id == -1 {
+		return nil, errors.New("no such voice")
+	}
+
+	// copy
+	res, err := http.Post(config.CurrentConfig.Voices.AivisSpeech.Api+"/audio_query?speaker="+strconv.Itoa(id)+"&text="+url.QueryEscape(content), "", nil)
+	if err != nil {
+		return nil, err
+	}
+	defer res.Body.Close()
+	if res.StatusCode != 200 {
+		return nil, errors.New("Response code error from aivisspeech:" + strconv.Itoa(res.StatusCode))
+	}
+
+	// copy
+	req := *voiceSource.Request
+
+	query := res.Body
+	buf := new(bytes.Buffer)
+	length, err := buf.ReadFrom(query)
+	if err != nil {
+		return nil, err
+	}
+	req.URL.RawQuery = "speaker=" + strconv.Itoa(id)
+	req.Body = io.NopCloser(buf)
+	req.ContentLength = length
+	req.GetBody = func() (io.ReadCloser, error) { return req.Body, nil }
+	req.Header.Set("Content-Type", "application/json")
+
+	res, err = httpCli.Do(&req)
+	if err != nil {
+		return nil, err
+	}
+	defer res.Body.Close()
+	bin, err := io.ReadAll(res.Body)
+	if err != nil {
+		return nil, err
+	}
+	if res.StatusCode != 200 {
+		return nil, errors.New("Response code error from aivisspeech:" + strconv.Itoa(res.StatusCode) + " " + string(bin))
+	}
+	return bin, nil
+}
+
+func (voiceSource aivisspeech) Verify(voice string) error {
+	for _, speaker := range voiceSource.Speakers.Speakers {
+		for _, v := range speaker.Styles {
+			if speaker.Name+v.Name == voice {
+				return nil
+			}
+		}
+	}
+	return errors.New("no such voice")
+}
+
+func (voiceSource aivisspeech) GetInfo() VoiceInfo {
+	return voiceSource.Info
+}

lib/voices/helper.go

@@ -2,16 +2,17 @@ package voices
 
 import (
 	"errors"
-	"github.com/tpc3/Kotone-DiVE/lib/config"
-	"github.com/tpc3/Kotone-DiVE/lib/db"
-	"github.com/tpc3/Kotone-DiVE/lib/utils"
 	"hash/crc64"
 	"io"
 	"log"
 	"net/http"
 	"strconv"
 	"time"
 
+	"github.com/tpc3/Kotone-DiVE/lib/config"
+	"github.com/tpc3/Kotone-DiVE/lib/db"
+	"github.com/tpc3/Kotone-DiVE/lib/utils"
+
 	"github.com/bwmarrin/discordgo"
 	"github.com/patrickmn/go-cache"
 )
@@ -63,6 +64,8 @@ func SourceSwitcher(source string) (VoiceSource, error) {
 		voiceSource = Voicevox
 	case Coeiroink.Info.Type:
 		voiceSource = Coeiroink
+	case AivisSpeech.Info.Type:
+		voiceSource = AivisSpeech
 	case AquestalkProxy.Info.Type:
 		voiceSource = AquestalkProxy
 	default:
@@ -102,7 +105,7 @@ func GetVoice(content string, voice *config.Voice) ([]byte, error) {
 		}
 
 		if bin == nil {
-			//Nothing to read
+			// Nothing to read
 			return nil, nil
 		}
 
@@ -139,7 +142,7 @@ func ReadVoice(session *discordgo.Session, orgMsg *discordgo.MessageCreate, enco
 	}
 
 	if session.VoiceConnections[orgMsg.GuildID] == nil {
-		return Skipped //Skipped due to the disconnection
+		return Skipped // Skipped due to the disconnection
 	}
 
 	frames, err := SplitToFrame(encoded)