Feat: google-translate-tts support

7 files changed, 208 insertions, 18 deletions

diff --git a/config.example.toml b/config.example.toml
index 2227c5e..eb1be70 100644
--- a/config.example.toml
+++ b/config.example.toml
@@ -29,6 +29,8 @@ RAGDir = "ragimport"
 TTS_ENABLED = false
 TTS_URL = "http://localhost:8880/v1/audio/speech"
 TTS_SPEED = 1.2
+TTS_PROVIDER = "kokoro"
+TTS_LANGUAGE = "en"
 # extra stt
 STT_ENABLED = false
 STT_TYPE = "WHISPER_SERVER" # WHISPER_SERVER or WHISPER_BINARY
diff --git a/config/config.go b/config/config.go
index 1454b4c..dc769d3 100644
--- a/config/config.go
+++ b/config/config.go
@@ -48,9 +48,11 @@ type Config struct {
 	OpenRouterToken         string `toml:"OpenRouterToken"`
 	OpenRouterModel         string `toml:"OpenRouterModel"`
 	// TTS
-	TTS_URL     string  `toml:"TTS_URL"`
-	TTS_ENABLED bool    `toml:"TTS_ENABLED"`
-	TTS_SPEED   float32 `toml:"TTS_SPEED"`
+	TTS_URL      string  `toml:"TTS_URL"`
+	TTS_ENABLED  bool    `toml:"TTS_ENABLED"`
+	TTS_SPEED    float32 `toml:"TTS_SPEED"`
+	TTS_PROVIDER string  `toml:"TTS_PROVIDER"`
+	TTS_LANGUAGE string  `toml:"TTS_LANGUAGE"`
 	// STT
 	STT_TYPE          string `toml:"STT_TYPE"` // WHISPER_SERVER, WHISPER_BINARY
 	STT_URL           string `toml:"STT_URL"`
diff --git a/docs/config.md b/docs/config.md
index fa5d42b..f1cac5b 100644
--- a/docs/config.md
+++ b/docs/config.md
@@ -96,11 +96,21 @@ This document explains how to set up and configure the application using the `co
 - Enable or disable text-to-speech functionality.
 
 #### TTS_URL (`"http://localhost:8880/v1/audio/speech"`)
-- The endpoint for TTS API.
+- The endpoint for TTS API (used with `kokoro` provider).
 
 #### TTS_SPEED (`1.2`)
 - Playback speed for speech output (1.0 is normal speed).
 
+#### TTS_PROVIDER (`"kokoro"`)
+- TTS provider to use. Options: `"kokoro"` or `"google"`.
+  - `"kokoro"`: Uses Kokoro FastAPI TTS server (requires TTS_URL to be set). Provides high-quality voice synthesis but requires a running Kokoro server.
+  - `"google"`: Uses Google Translate TTS with gopxl/beep for local playback. Works offline using Google's public TTS API with local audio playback via gopxl/beep. Supports multiple languages via TTS_LANGUAGE setting.
+
+#### TTS_LANGUAGE (`"en"`)
+- Language code for TTS (used with `google` provider).
+  - Examples: `"en"` (English), `"es"` (Spanish), `"fr"` (French)
+  - See Google Translate TTS documentation for supported languages.
+
 ### Speech-to-Text (STT) Settings
 
 #### STT_ENABLED (`false`)
diff --git a/docs/tutorial_rp.md b/docs/tutorial_rp.md
index 451aadb..9053ffb 100644
--- a/docs/tutorial_rp.md
+++ b/docs/tutorial_rp.md
@@ -1,5 +1,7 @@
-after [installing](https://github.com/GrailFinder/gf-lt/tree/master?tab=readme-ov-file#how-to-install)
-[set up your config](config.md)
+### RP case example
+
+check the (https://github.com/GrailFinder/gf-lt/tree/master?tab=readme-ov-file#how-to-install) and
+[setting up your config](config.md)
 
 To roleplay, we would need to create a character card or get one from the web.
 For this tutorial, we are going to use the default character Seraphina from [SillyTavern (ST)](https://github.com/SillyTavern/SillyTavern/blob/release/default/content/default_Seraphina.png).
diff --git a/extra/tts.go b/extra/tts.go
index c9ad59d..0209072 100644
--- a/extra/tts.go
+++ b/extra/tts.go
@@ -12,9 +12,12 @@ import (
 	"io"
 	"log/slog"
 	"net/http"
+	"os"
 	"strings"
 	"time"
 
+	google_translate_tts "github.com/GrailFinder/google-translate-tts"
+	"github.com/GrailFinder/google-translate-tts/handlers"
 	"github.com/gopxl/beep/v2"
 	"github.com/gopxl/beep/v2/mp3"
 	"github.com/gopxl/beep/v2/speaker"
@@ -49,6 +52,14 @@ type KokoroOrator struct {
 	// textBuffer bytes.Buffer
 }
 
+// Google Translate TTS implementation
+type GoogleTranslateOrator struct {
+	logger        *slog.Logger
+	speech        *google_translate_tts.Speech
+	currentStream *beep.Ctrl
+	textBuffer    strings.Builder
+}
+
 func (o *KokoroOrator) stoproutine() {
 	<-TTSDoneChan
 	o.logger.Debug("orator got done signal")
@@ -123,18 +134,47 @@ func (o *KokoroOrator) readroutine() {
 }
 
 func NewOrator(log *slog.Logger, cfg *config.Config) Orator {
-	orator := &KokoroOrator{
-		logger:   log,
-		URL:      cfg.TTS_URL,
-		Format:   models.AFMP3,
-		Stream:   false,
-		Speed:    cfg.TTS_SPEED,
-		Language: "a",
-		Voice:    "af_bella(1)+af_sky(1)",
-	}
-	go orator.readroutine()
-	go orator.stoproutine()
-	return orator
+	provider := cfg.TTS_PROVIDER
+	if provider == "" {
+		provider = "kokoro"
+	}
+
+	switch strings.ToLower(provider) {
+	case "google", "google-translate", "google_translate":
+		language := cfg.TTS_LANGUAGE
+		if language == "" {
+			language = "en"
+		}
+
+		speech := &google_translate_tts.Speech{
+			Folder:   os.TempDir() + "/gf-lt-tts", // Temporary directory for caching
+			Language: language,
+			Proxy:    "", // Proxy not supported
+			Speed:    cfg.TTS_SPEED,
+			Handler:  &handlers.Beep{},
+		}
+
+		orator := &GoogleTranslateOrator{
+			logger: log,
+			speech: speech,
+		}
+		go orator.readroutine()
+		go orator.stoproutine()
+		return orator
+	default: // kokoro
+		orator := &KokoroOrator{
+			logger:   log,
+			URL:      cfg.TTS_URL,
+			Format:   models.AFMP3,
+			Stream:   false,
+			Speed:    cfg.TTS_SPEED,
+			Language: "a",
+			Voice:    "af_bella(1)+af_sky(1)",
+		}
+		go orator.readroutine()
+		go orator.stoproutine()
+		return orator
+	}
 }
 
 func (o *KokoroOrator) GetLogger() *slog.Logger {
@@ -213,3 +253,132 @@ func (o *KokoroOrator) Stop() {
 		o.currentStream.Streamer = nil
 	}
 }
+
+func (o *GoogleTranslateOrator) stoproutine() {
+	<-TTSDoneChan
+	o.logger.Debug("orator got done signal")
+	o.Stop()
+	// drain the channel
+	for len(TTSTextChan) > 0 {
+		<-TTSTextChan
+	}
+}
+
+func (o *GoogleTranslateOrator) readroutine() {
+	tokenizer, _ := english.NewSentenceTokenizer(nil)
+	for {
+		select {
+		case chunk := <-TTSTextChan:
+			_, err := o.textBuffer.WriteString(chunk)
+			if err != nil {
+				o.logger.Warn("failed to write to stringbuilder", "error", err)
+				continue
+			}
+			text := o.textBuffer.String()
+			sentences := tokenizer.Tokenize(text)
+			o.logger.Debug("adding chunk", "chunk", chunk, "text", text, "sen-len", len(sentences))
+			for i, sentence := range sentences {
+				if i == len(sentences)-1 { // last sentence
+					o.textBuffer.Reset()
+					_, err := o.textBuffer.WriteString(sentence.Text)
+					if err != nil {
+						o.logger.Warn("failed to write to stringbuilder", "error", err)
+						continue
+					}
+					continue // if only one (often incomplete) sentence; wait for next chunk
+				}
+				o.logger.Debug("calling Speak with sentence", "sent", sentence.Text)
+				if err := o.Speak(sentence.Text); err != nil {
+					o.logger.Error("tts failed", "sentence", sentence.Text, "error", err)
+				}
+			}
+		case <-TTSFlushChan:
+			o.logger.Debug("got flushchan signal start")
+			// lln is done get the whole message out
+			if len(TTSTextChan) > 0 { // otherwise might get stuck
+				for chunk := range TTSTextChan {
+					_, err := o.textBuffer.WriteString(chunk)
+					if err != nil {
+						o.logger.Warn("failed to write to stringbuilder", "error", err)
+						continue
+					}
+					if len(TTSTextChan) == 0 {
+						break
+					}
+				}
+			}
+			// INFO: if there is a lot of text it will take some time to make with tts at once
+			// to avoid this pause, it might be better to keep splitting on sentences
+			// but keepinig in mind that remainder could be ommited by tokenizer
+			// Flush remaining text
+			remaining := o.textBuffer.String()
+			o.textBuffer.Reset()
+			if remaining != "" {
+				o.logger.Debug("calling Speak with remainder", "rem", remaining)
+				if err := o.Speak(remaining); err != nil {
+					o.logger.Error("tts failed", "sentence", remaining, "error", err)
+				}
+			}
+		}
+	}
+}
+
+func (o *GoogleTranslateOrator) GetLogger() *slog.Logger {
+	return o.logger
+}
+
+func (o *GoogleTranslateOrator) Speak(text string) error {
+	o.logger.Debug("fn: Speak is called", "text-len", len(text))
+
+	// Generate MP3 data using google-translate-tts
+	reader, err := o.speech.GenerateSpeech(text)
+	if err != nil {
+		o.logger.Error("generate speech failed", "error", err)
+		return fmt.Errorf("generate speech failed: %w", err)
+	}
+
+	// Decode the mp3 audio from reader (wrap with NopCloser for io.ReadCloser)
+	streamer, format, err := mp3.Decode(io.NopCloser(reader))
+	if err != nil {
+		o.logger.Error("mp3 decode failed", "error", err)
+		return fmt.Errorf("mp3 decode failed: %w", err)
+	}
+	defer streamer.Close()
+
+	playbackStreamer := beep.Streamer(streamer)
+	speed := o.speech.Speed
+	if speed <= 0 {
+		speed = 1.0
+	}
+	if speed != 1.0 {
+		playbackStreamer = beep.ResampleRatio(3, float64(speed), streamer)
+	}
+
+	// Initialize speaker with the format's sample rate
+	if err := speaker.Init(format.SampleRate, format.SampleRate.N(time.Second/10)); err != nil {
+		o.logger.Debug("failed to init speaker", "error", err)
+	}
+
+	done := make(chan bool)
+	// Create controllable stream and store reference
+	o.currentStream = &beep.Ctrl{Streamer: beep.Seq(playbackStreamer, beep.Callback(func() {
+		close(done)
+		o.currentStream = nil
+	})), Paused: false}
+	speaker.Play(o.currentStream)
+	<-done // wait for playback to complete
+	return nil
+}
+
+func (o *GoogleTranslateOrator) Stop() {
+	o.logger.Debug("attempted to stop google translate orator")
+	speaker.Lock()
+	defer speaker.Unlock()
+	if o.currentStream != nil {
+		o.currentStream.Streamer = nil
+	}
+	// Also stop the speech handler if possible
+	if o.speech != nil {
+		_ = o.speech.Stop()
+	}
+}
diff --git a/go.mod b/go.mod
index 82a7039..944a920 100644
--- a/go.mod
+++ b/go.mod
@@ -4,6 +4,7 @@ go 1.25.1
 
 require (
 	github.com/BurntSushi/toml v1.5.0
+	github.com/GrailFinder/google-translate-tts v0.0.0-00010101000000-000000000000
 	github.com/GrailFinder/searchagent v0.2.0
 	github.com/gdamore/tcell/v2 v2.13.2
 	github.com/glebarez/go-sqlite v1.22.0
@@ -23,6 +24,7 @@ require (
 	github.com/gdamore/encoding v1.0.1 // indirect
 	github.com/google/uuid v1.6.0 // indirect
 	github.com/hajimehoshi/go-mp3 v0.3.4 // indirect
+	github.com/hajimehoshi/oto/v2 v2.3.1 // indirect
 	github.com/lucasb-eyer/go-colorful v1.3.0 // indirect
 	github.com/mattn/go-isatty v0.0.20 // indirect
 	github.com/ncruces/go-strftime v1.0.0 // indirect
@@ -39,3 +41,5 @@ require (
 	modernc.org/memory v1.11.0 // indirect
 	modernc.org/sqlite v1.40.1 // indirect
 )
+
+replace github.com/GrailFinder/google-translate-tts => /home/grail/projects/plays/goplays/google-translate-tts
diff --git a/go.sum b/go.sum
index 465b449..6857a00 100644
--- a/go.sum
+++ b/go.sum
@@ -35,6 +35,7 @@ github.com/gordonklaus/portaudio v0.0.0-20250206071425-98a94950218b h1:WEuQWBxel
 github.com/gordonklaus/portaudio v0.0.0-20250206071425-98a94950218b/go.mod h1:esZFQEUwqC+l76f2R8bIWSwXMaPbp79PppwZ1eJhFco=
 github.com/hajimehoshi/go-mp3 v0.3.4 h1:NUP7pBYH8OguP4diaTZ9wJbUbk3tC0KlfzsEpWmYj68=
 github.com/hajimehoshi/go-mp3 v0.3.4/go.mod h1:fRtZraRFcWb0pu7ok0LqyFhCUrPeMsGRSVop0eemFmo=
+github.com/hajimehoshi/oto/v2 v2.3.1 h1:qrLKpNus2UfD674oxckKjNJmesp9hMh7u7QCrStB3Rc=
 github.com/hajimehoshi/oto/v2 v2.3.1/go.mod h1:seWLbgHH7AyUMYKfKYT9pg7PhUu9/SisyJvNTT+ASQo=
 github.com/hashicorp/golang-lru/v2 v2.0.7 h1:a+bsQ5rvGLjzHuww6tVxozPZFVghXaHOwFs4luLUK2k=
 github.com/hashicorp/golang-lru/v2 v2.0.7/go.mod h1:QeFd9opnmA6QUJc5vARoKUSoFhyfM2/ZepoAG6RGpeM=