M src/app/proxy.go → src/app/proxy.go

@@ -103,13 +103,7 @@ defer res.Body.Close()
 
 	w.Header().Set("Content-Type", "text/vtt")
 
-	content, err := io.ReadAll(res.Body)
-	if err != nil {
-		http.Error(w, err500, http.StatusInternalServerError)
-		return
-	}
-
-	err = subs.Parse(content, w)
+	err = subs.Convert(res.Body, w)
 	if err != nil {
 		http.Error(w, err500, http.StatusInternalServerError)
 		return

M src/subs/subs.go → src/subs/subs.go

@@ -3,8 +3,9 @@
 import (
 	"encoding/xml"
 	"fmt"
+	"io"
 	"log"
-	"net/http"
+	"strings"
 )
 
 type TimedText struct {
@@ -17,9 +18,28 @@ Paragraphs []Paragraph `xml:"p"`
 }
 
 type Paragraph struct {
-	Text   string `xml:",chardata"`
-	Start  int    `xml:"t,attr"` // Start time in milliseconds
-	Length int    `xml:"d,attr"` // Duration in milliseconds
+	Start     int        `xml:"t,attr"`    // Start time in milliseconds
+	Length    int        `xml:"d,attr"`    // Duration in milliseconds
+	Text      string     `xml:",chardata"` // Direct text (for cases without <s> tags)
+	Sentences []Sentence `xml:"s"`         // List of <s> tags (for cases with individual words/phrases)
+}
+
+type Sentence struct {
+	Text string `xml:",chardata"` // Text inside the <s> tag
+	Time int    `xml:"t,attr"`    // Optional start time (not always present)
+}
+
+func writeVTT(output io.Writer, i, startTime, endTime int, sentence string) {
+	output.Write(
+		[]byte(
+			fmt.Sprintf(
+				"%d\n%s --> %s\n%s\n\n",
+				i,
+				millisecondsToTimestamp(startTime),
+				millisecondsToTimestamp(endTime),
+				sentence,
+			),
+		))
 }
 
 // Convert milliseconds to WebVTT timestamp format: HH:MM:SS.mmm
@@ -29,28 +49,46 @@ milliseconds := ms % 1000
 	return fmt.Sprintf("%02d:%02d:%02d.%03d", seconds/3600, (seconds%3600)/60, seconds%60, milliseconds)
 }
 
-func writeString(s string, w http.ResponseWriter) {
-	w.Write([]byte(s))
-}
+func Convert(reader io.Reader, output io.Writer) error {
+	content, err := io.ReadAll(reader)
+	if err != nil {
+		return err
+	}
 
-func Parse(inputData []byte, w http.ResponseWriter) error {
 	var timedText TimedText
-	err := xml.Unmarshal(inputData, &timedText)
+	err = xml.Unmarshal(content, &timedText)
 	if err != nil {
 		log.Println("Error unmarshalling XML:", err)
 		return err
 	}
 
-	// Write the WebVTT header
-	writeString("WEBVTT\n\n", w)
+	output.Write([]byte("WEBVTT\n\n"))
 
-	// Loop through the paragraphs and write them to the WebVTT file
+	var lastEndTime int
 	for i, p := range timedText.Body.Paragraphs {
-		startTime := millisecondsToTimestamp(p.Start)
-		endTime := millisecondsToTimestamp(p.Start + p.Length)
+		startTimeMS := p.Start
+		endTimeMS := p.Start + p.Length
+
+		if startTimeMS < lastEndTime {
+			startTimeMS = lastEndTime
+		}
+
+		var sentence string
+		if len(p.Sentences) > 0 {
+			for _, s := range p.Sentences {
+				sentence += s.Text
+			}
+		} else {
+			sentence = p.Text
+		}
+
+		sentence = strings.TrimSpace(sentence)
+		if sentence == "" {
+			continue
+		}
 
-		s := fmt.Sprintf("%d\n%s --> %s\n%s\n\n", i+1, startTime, endTime, p.Text)
-		writeString(s, w)
+		lastEndTime = endTimeMS
+		writeVTT(output, i+1, startTimeMS, endTimeMS, sentence)
 	}
 
 	return nil