chrome tts

This commit is contained in:
legop3
2026-06-01 00:56:01 -04:00
parent e108eaa4eb
commit ac93d3227b
15 changed files with 588 additions and 56 deletions
+161
View File
@@ -0,0 +1,161 @@
package roverd
import (
"bufio"
"context"
"encoding/json"
"fmt"
"io"
"log"
"os"
"os/exec"
"sync"
)
const chromeTTSDaemonPath = "/usr/local/bin/chromegtts-daemon"
type chromeTTSDaemon struct {
log *log.Logger
mu sync.Mutex
cmd *exec.Cmd
stdin io.WriteCloser
scanner *bufio.Scanner
}
func NewChromeTTSDaemon(logger *log.Logger) *chromeTTSDaemon {
return &chromeTTSDaemon{log: logger}
}
func (d *chromeTTSDaemon) Start(ctx context.Context) error {
d.mu.Lock()
defer d.mu.Unlock()
return d.startLocked(ctx)
}
func (d *chromeTTSDaemon) Speak(ctx context.Context, text, voice string, pitch, speed float64) error {
d.mu.Lock()
defer d.mu.Unlock()
if err := d.startLocked(ctx); err != nil {
return err
}
payload := map[string]any{
"text": text,
"voice": voice,
"pitch": pitch,
"speed": speed,
}
encoded, err := json.Marshal(payload)
if err != nil {
return err
}
if _, err := d.stdin.Write(append(encoded, '\n')); err != nil {
d.stopLocked()
return fmt.Errorf("write chromegtts request: %w", err)
}
type result struct {
response chromeTTSResponse
err error
}
done := make(chan result, 1)
go func(scanner *bufio.Scanner) {
var response chromeTTSResponse
if !scanner.Scan() {
done <- result{err: fmt.Errorf("chromegtts daemon stopped")}
return
}
if err := json.Unmarshal(scanner.Bytes(), &response); err != nil {
done <- result{err: fmt.Errorf("decode chromegtts response: %w", err)}
return
}
done <- result{response: response}
}(d.scanner)
select {
case <-ctx.Done():
d.stopLocked()
return ctx.Err()
case res := <-done:
if res.err != nil {
d.stopLocked()
return res.err
}
if !res.response.OK {
return fmt.Errorf("chromegtts failed: %s", res.response.Error)
}
return nil
}
}
func (d *chromeTTSDaemon) Shutdown() {
d.mu.Lock()
defer d.mu.Unlock()
d.stopLocked()
}
type chromeTTSResponse struct {
OK bool `json:"ok"`
Ready bool `json:"ready,omitempty"`
Error string `json:"error,omitempty"`
}
func (d *chromeTTSDaemon) startLocked(ctx context.Context) error {
if d.cmd != nil && d.cmd.ProcessState == nil {
return nil
}
cmd := exec.Command(chromeTTSDaemonPath)
stdin, err := cmd.StdinPipe()
if err != nil {
return err
}
stdout, err := cmd.StdoutPipe()
if err != nil {
return err
}
cmd.Stderr = os.Stderr
if err := cmd.Start(); err != nil {
return fmt.Errorf("start chromegtts daemon: %w", err)
}
scanner := bufio.NewScanner(stdout)
var ready chromeTTSResponse
if !scanner.Scan() {
_ = cmd.Process.Kill()
_ = cmd.Wait()
return fmt.Errorf("chromegtts daemon exited before ready")
}
if err := json.Unmarshal(scanner.Bytes(), &ready); err != nil {
_ = cmd.Process.Kill()
_ = cmd.Wait()
return fmt.Errorf("decode chromegtts ready: %w", err)
}
if !ready.OK {
_ = cmd.Process.Kill()
_ = cmd.Wait()
return fmt.Errorf("chromegtts daemon not ready: %s", ready.Error)
}
d.cmd = cmd
d.stdin = stdin
d.scanner = scanner
if d.log != nil {
d.log.Printf("chromegtts daemon started")
}
return nil
}
func (d *chromeTTSDaemon) stopLocked() {
if d.stdin != nil {
_ = d.stdin.Close()
d.stdin = nil
}
if d.cmd != nil && d.cmd.ProcessState == nil && d.cmd.Process != nil {
_ = d.cmd.Process.Kill()
_ = d.cmd.Wait()
}
d.cmd = nil
d.scanner = nil
}
+6 -5
View File
@@ -64,11 +64,12 @@ type servoPayload struct {
}
type ttsPayload struct {
Text string `json:"text"`
Engine string `json:"engine,omitempty"`
Voice string `json:"voice,omitempty"`
Pitch int `json:"pitch,omitempty"`
Speak bool `json:"speak,omitempty"`
Text string `json:"text"`
Engine string `json:"engine,omitempty"`
Voice string `json:"voice,omitempty"`
Pitch float64 `json:"pitch,omitempty"`
Speed float64 `json:"speed,omitempty"`
Speak bool `json:"speak,omitempty"`
}
type hornPayload struct {
+11 -6
View File
@@ -38,11 +38,11 @@ func (c *WSClient) handleTTSPayload(ctx context.Context, payload *ttsPayload) er
if voice == "" {
voice = strings.TrimSpace(c.cfg.Audio.DefaultVoice)
}
pitch := payload.Pitch
if pitch <= 0 {
pitch = c.cfg.Audio.DefaultPitch
espeakPitch := int(payload.Pitch)
if espeakPitch <= 0 {
espeakPitch = c.cfg.Audio.DefaultPitch
}
pitch = clampInt(pitch, 0, 99)
espeakPitch = clampInt(espeakPitch, 0, 99)
runCtx, cancel := context.WithTimeout(ctx, 12*time.Second)
defer cancel()
@@ -51,8 +51,8 @@ func (c *WSClient) handleTTSPayload(ctx context.Context, payload *ttsPayload) er
switch engine {
case "espeak", "e":
args := []string{}
if pitch > 0 {
args = append(args, "-p", fmt.Sprintf("%d", pitch))
if espeakPitch > 0 {
args = append(args, "-p", fmt.Sprintf("%d", espeakPitch))
}
args = append(args, text)
cmd = exec.CommandContext(runCtx, "espeak", args...)
@@ -63,6 +63,11 @@ func (c *WSClient) handleTTSPayload(ctx context.Context, payload *ttsPayload) er
}
args = append(args, "-t", text)
cmd = exec.CommandContext(runCtx, "flite", args...)
case "chromegtts", "googletts", "gtts", "google":
if c.chromeTTS == nil {
return fmt.Errorf("chromegtts unavailable")
}
return c.chromeTTS.Speak(runCtx, text, voice, payload.Pitch, payload.Speed)
default:
return fmt.Errorf("unsupported tts engine: %s", engine)
}
+21
View File
@@ -26,6 +26,7 @@ type WSClient struct {
recoverMu sync.Mutex
recovering bool
ttsQueue chan *ttsPayload
chromeTTS *chromeTTSDaemon
lastAux motorPWMPayload
autoSideOn bool
connMu sync.Mutex
@@ -43,6 +44,10 @@ func NewWSClient(cfg *Config, adapter *SerialAdapter, frames <-chan []byte, even
if cfg.Audio.TTSEnabled {
ttsQueue = make(chan *ttsPayload, 2)
}
var chromeTTS *chromeTTSDaemon
if cfg.Audio.TTSEnabled {
chromeTTS = NewChromeTTSDaemon(logger)
}
var horn *HornSynth
if cfg.Horn.Enabled {
horn = NewHornSynth(cfg.Horn, logger)
@@ -58,6 +63,7 @@ func NewWSClient(cfg *Config, adapter *SerialAdapter, frames <-chan []byte, even
nightVision: nightVision,
log: logger,
ttsQueue: ttsQueue,
chromeTTS: chromeTTS,
audioLevels: AudioLevels{
HornGain: 1.0,
TTSGain: 1.0,
@@ -79,6 +85,9 @@ func (c *WSClient) Run(ctx context.Context) error {
c.markConnected()
defer conn.Close(websocket.StatusInternalError, "closed")
defer c.markDisconnected()
if c.chromeTTS != nil {
defer c.chromeTTS.Shutdown()
}
if err := c.sendHello(ctx, conn); err != nil {
return err
@@ -89,6 +98,7 @@ func (c *WSClient) Run(ctx context.Context) error {
errCh := make(chan error, 2)
c.startTTSWorker(ctx)
c.warmChromeTTS(ctx)
go func() {
errCh <- c.readLoop(ctx, conn)
}()
@@ -356,6 +366,17 @@ func (c *WSClient) startTTSWorker(ctx context.Context) {
}()
}
func (c *WSClient) warmChromeTTS(ctx context.Context) {
if c.chromeTTS == nil {
return
}
go func() {
if err := c.chromeTTS.Start(ctx); err != nil {
c.log.Printf("chromegtts warmup failed: %v", err)
}
}()
}
func (c *WSClient) handleServoCommand(payload *servoPayload) error {
switch {
case payload.Angle != nil: