From 605a6b2b280c0b2d555f0e2fad56b289361ce29c Mon Sep 17 00:00:00 2001
From: Ariver <shanghai3168@gmail.com>
Date: Sun, 07 Jun 2026 11:28:52 +0800
Subject: [PATCH] Add X-ASR live captions
---
privatevoice.src/app.go | 129 ++++++++++++++++++++++++++++++++++++++++--
1 files changed, 121 insertions(+), 8 deletions(-)
diff --git a/privatevoice.src/app.go b/privatevoice.src/app.go
index d4e1ccd..ad5b0e4 100755
--- a/privatevoice.src/app.go
+++ b/privatevoice.src/app.go
@@ -3,6 +3,7 @@
import (
"context"
"fmt"
+ "strings"
"sync"
"sync/atomic"
"time"
@@ -27,8 +28,8 @@
)
const (
- appVersion = "2.1.29"
- appBuild = "20260607.0147"
+ appVersion = "2.1.30"
+ appBuild = "20260607.1127"
appDisplayVersion = appVersion + " (build " + appBuild + ")"
appName = "PrivateVoice Dictation"
@@ -37,6 +38,8 @@
silenceGracePeriod = 2 * time.Second // don't auto-stop within first 2s of free-talk
holdActivationDelay = 180 * time.Millisecond
doneIndicatorHideDelayMs = 250
+ liveCaptionInterval = 120 * time.Millisecond
+ liveCaptionMaxBytes = 108
)
// App holds all application state and orchestration logic.
@@ -75,6 +78,8 @@
lastHotkeyBlocked time.Time
silenceSince time.Time // when continuous silence started (free-talk only)
freeTalkStart time.Time // when free-talk mode started
+ liveCaptionCancel context.CancelFunc
+ liveCaptionSeq uint64
}
func RunApp() error {
@@ -421,6 +426,7 @@
a.isRecording = false
return
}
+ a.startLiveCaptionLocked(overlay.StatusRecording)
a.startRecordingTimer(overlay.StatusRecording)
}
@@ -433,6 +439,7 @@
if cancel {
logger.Info("Recording cancelled (combination key)")
+ a.stopLiveCaptionLocked()
a.indicator.SetStatus(overlay.StatusCancelled, "已取消")
a.recorder.Stop()
if a.cfg.SoundFeedback {
@@ -442,6 +449,7 @@
return
}
+ a.stopLiveCaptionLocked()
hasVoice := a.recorder.HasVoiceActivity()
a.indicator.SetStatus(overlay.StatusProcessing, "识别中")
go func() {
@@ -474,6 +482,7 @@
a.isRecording = false
return
}
+ a.startLiveCaptionLocked(overlay.StatusFreetalking)
a.startRecordingTimer(overlay.StatusFreetalking)
}
@@ -491,15 +500,120 @@
case <-ticker.C:
a.mu.Lock()
recording := a.isRecording
+ captionActive := a.liveCaptionCancel != nil
a.mu.Unlock()
if !recording {
return
+ }
+ if captionActive {
+ continue
}
d := time.Since(start)
a.indicator.SetStatus(status, fmt.Sprintf("%d:%02d", int(d.Minutes()), int(d.Seconds())%60))
}
}
}()
+}
+
+func (a *App) startLiveCaptionLocked(status overlay.Status) {
+ if a.liveCaptionCancel != nil {
+ return
+ }
+
+ a.engineMu.Lock()
+ streamingEng, ok := a.eng.(engine.StreamingEngine)
+ if !ok {
+ a.engineMu.Unlock()
+ return
+ }
+
+ ctx, cancel := context.WithCancel(a.ctx)
+ a.liveCaptionSeq++
+ seq := a.liveCaptionSeq
+ a.liveCaptionCancel = cancel
+ go a.runLiveCaption(ctx, seq, status, streamingEng, a.engineMu.Unlock)
+}
+
+func (a *App) stopLiveCaptionLocked() {
+ if a.liveCaptionCancel == nil {
+ return
+ }
+ a.liveCaptionCancel()
+ a.liveCaptionCancel = nil
+ a.liveCaptionSeq++
+}
+
+func (a *App) clearLiveCaption(seq uint64) {
+ a.mu.Lock()
+ defer a.mu.Unlock()
+ if a.liveCaptionSeq == seq {
+ a.liveCaptionCancel = nil
+ }
+}
+
+func (a *App) runLiveCaption(ctx context.Context, seq uint64, status overlay.Status, streamingEng engine.StreamingEngine, releaseEngine func()) {
+ defer a.clearLiveCaption(seq)
+ defer releaseEngine()
+
+ session, err := streamingEng.NewStreamingSession()
+ if err != nil {
+ logger.Error("Live caption disabled: %v", err)
+ return
+ }
+ defer session.Close()
+
+ ticker := time.NewTicker(liveCaptionInterval)
+ defer ticker.Stop()
+
+ offset := 0
+ lastDisplay := ""
+ for {
+ select {
+ case <-ctx.Done():
+ return
+ case <-ticker.C:
+ samples, nextOffset := a.recorder.ReadSamplesSince(offset)
+ offset = nextOffset
+ if len(samples) == 0 {
+ continue
+ }
+ partial, err := session.Accept(samples)
+ if err != nil {
+ logger.Error("Live caption failed: %v", err)
+ return
+ }
+ display := liveCaptionDisplayText(a.userdict.Apply(textproc.PostProcess(partial)))
+ if display == "" || display == lastDisplay {
+ continue
+ }
+
+ a.mu.Lock()
+ active := a.isRecording && a.liveCaptionSeq == seq
+ a.mu.Unlock()
+ if !active {
+ return
+ }
+
+ lastDisplay = display
+ a.indicator.SetStatus(status, display)
+ }
+ }
+}
+
+func liveCaptionDisplayText(text string) string {
+ text = strings.TrimSpace(text)
+ if text == "" || len([]byte(text)) <= liveCaptionMaxBytes {
+ return text
+ }
+
+ runes := []rune(text)
+ for len(runes) > 0 && len([]byte("..."+string(runes))) > liveCaptionMaxBytes {
+ runes = runes[1:]
+ }
+ if len(runes) == 0 {
+ return ""
+ }
+ return "..." + string(runes)
}
func (a *App) stopTapRecordingLocked() {
@@ -510,6 +624,7 @@
a.isRecording = false
a.lastStopTime = time.Now()
+ a.stopLiveCaptionLocked()
hasVoice := a.recorder.HasVoiceActivity()
logger.Info("Tap recording stopped")
a.indicator.SetStatus(overlay.StatusProcessing, "识别中")
@@ -556,6 +671,7 @@
a.isFreetalking = false
a.isRecording = false
a.lastStopTime = time.Now()
+ a.stopLiveCaptionLocked()
// MUST stop device in a separate goroutine: we are inside the audio
// data callback, and device.Stop() waits for in-flight callbacks to
// finish — calling it here would deadlock.
@@ -742,12 +858,9 @@
func (a *App) replaceEngine(eng engine.Engine) {
a.engineMu.Lock()
- defer a.engineMu.Unlock()
-
- a.mu.Lock()
old := a.eng
a.eng = eng
- a.mu.Unlock()
+ a.engineMu.Unlock()
if old != nil && old != eng {
old.Close()
@@ -961,8 +1074,8 @@
// GetEngine returns the current engine (may be nil).
func (a *App) GetEngine() engine.Engine {
- a.mu.Lock()
- defer a.mu.Unlock()
+ a.engineMu.Lock()
+ defer a.engineMu.Unlock()
return a.eng
}
--
Gitblit v1.9.3