first successful export

transcode/transcode.go

@@ -8,6 +8,7 @@ import (
 	"os/exec"
 	"regexp"
 	"strconv"
+	"strings"
 )
 
 // Segment is a time interval [Start, End) in seconds.
@@ -24,8 +25,8 @@ type wsMsg struct {
 	Message  string    `json:"message,omitempty"`
 }
 
-func send(msg wsMsg, broadcast func([]byte)) {
-	data, _ := json.Marshal(msg)
+func send(msg *wsMsg, broadcast func([]byte)) {
+	data, _ := json.Marshal(&msg)
 	broadcast(data)
 }
 
@@ -36,11 +37,7 @@ func send(msg wsMsg, broadcast func([]byte)) {
 // DetectSpeechSegments runs ffmpeg's silencedetect filter and returns the
 // non-silent (speech) segments. A "segments" WebSocket message is broadcast
 // when detection finishes.
-func DetectSpeechSegments(
-	inputPath string,
-	noiseDb, minDuration, padding float64,
-	broadcast func([]byte),
-) ([]Segment, error) {
+func DetectSpeechSegments(inputPath string, noiseDb, minDuration, padding float64, broadcast func([]byte)) ([]Segment, error) {
 	filter := fmt.Sprintf("silencedetect=noise=%.0fdB:d=%.2f", noiseDb, minDuration)
 
 	cmd := exec.Command("ffmpeg",
@@ -91,9 +88,9 @@ func DetectSpeechSegments(
 		return nil, fmt.Errorf("ffmpeg: %w", err)
 	}
 
-	segments := invertSilences(silences, totalDuration, padding)
+	segments := removeSilence(silences, totalDuration, padding)
 
-	send(wsMsg{
+	send(&wsMsg{
 		Type:     "segments",
 		Segments: segments,
 		Duration: totalDuration,
@@ -104,9 +101,9 @@ func DetectSpeechSegments(
 
 type silenceInterval struct{ start, end float64 }
 
-// invertSilences turns silence regions into speech regions, with a small
+// removeSilence turns silence regions into speech regions, with a small
 // padding buffer so words at the edges don't get clipped.
-func invertSilences(silences []silenceInterval, totalDuration, padding float64) []Segment {
+func removeSilence(silences []silenceInterval, totalDuration, padding float64) []Segment {
 	if len(silences) == 0 {
 		return []Segment{{Start: 0, End: totalDuration}}
 	}
@@ -136,37 +133,41 @@ func invertSilences(silences []silenceInterval, totalDuration, padding float64)
 	return segments
 }
 
-// ──────────────────────────────────────────────────────────────────────────────
-// Export
-// ──────────────────────────────────────────────────────────────────────────────
-
-// ExportSegments concatenates the given segments using ffmpeg's concat demuxer
-// (stream-copy, no re-encode). Progress is streamed via broadcast.
 func ExportSegments(
 	inputPath, outputPath string,
 	segments []Segment,
 	broadcast func([]byte),
 ) error {
-	concatPath := outputPath + ".concat.txt"
-	f, err := os.Create(concatPath)
-	if err != nil {
-		return fmt.Errorf("create concat file: %w", err)
-	}
-	defer os.Remove(concatPath)
-
 	totalDuration := 0.0
 	for _, seg := range segments {
 		totalDuration += seg.End - seg.Start
-		fmt.Fprintf(f, "file '%s'\ninpoint %.6f\noutpoint %.6f\n",
-			inputPath, seg.Start, seg.End)
 	}
-	f.Close()
+
+	// Build a filter_complex that trims each segment and resets timestamps,
+	// then concatenates them. This avoids non-monotonic DTS issues that occur
+	// when stream-copying segments with their original timestamps.
+	var filterParts []string
+	var concatInputs string
+	for i, seg := range segments {
+		filterParts = append(filterParts,
+			fmt.Sprintf("[0:v]trim=start=%.6f:end=%.6f,setpts=PTS-STARTPTS[v%d]", seg.Start, seg.End, i),
+			fmt.Sprintf("[0:a]atrim=start=%.6f:end=%.6f,asetpts=PTS-STARTPTS[a%d]", seg.Start, seg.End, i),
+		)
+		concatInputs += fmt.Sprintf("[v%d][a%d]", i, i)
+	}
+	filterParts = append(filterParts,
+		fmt.Sprintf("%sconcat=n=%d:v=1:a=1[outv][outa]", concatInputs, len(segments)),
+	)
+	filterComplex := strings.Join(filterParts, ";")
 
 	cmd := exec.Command("ffmpeg",
-		"-f", "concat",
-		"-safe", "0",
-		"-i", concatPath,
-		"-c", "copy",
+		"-i", inputPath,
+		"-filter_complex", filterComplex,
+		"-map", "[outv]",
+		"-map", "[outa]",
+		// around 80% compression
+		"-c:v", "libx264", "-crf", "28", "-preset", "fast",
+		"-c:a", "aac", "-b:a", "128k",
 		"-progress", "pipe:1",
 		"-y",
 		outputPath,
@@ -194,7 +195,7 @@ func ExportSegments(
 			if pct > 100 {
 				pct = 100
 			}
-			send(wsMsg{Type: "progress", Percent: pct}, broadcast)
+			send(&wsMsg{Type: "progress", Percent: pct}, broadcast)
 		}
 	}