Transcript: Use Whisper response without LF. (#163). v5.14.5

ossrs · Mar 10, 2024 · d051bf2 · d051bf2
1 parent 6b62ea6
commit d051bf2
Show file tree

Hide file tree

Showing 3 changed files with 18 additions and 20 deletions.
diff --git a/DEVELOPER.md b/DEVELOPER.md
@@ -1144,6 +1144,7 @@ The following are the update records for the SRS Stack server.
     * HLS: Set m3u8 expire time to 1s for LLHLS. v5.14.4
     * HLS: Use fast cache for HLS config. v5.14.4
     * Transcript: Support set the force_style for overlay subtitle. v5.14.5
+    * Transcript: Use Whisper response without LF. (#163). v5.14.5
 * v5.13:
     * Fix bug for vlive and transcript. v5.13.1
     * Support AWS Lightsail install script. v5.13.2
@@ -1178,6 +1179,7 @@ The following are the update records for the SRS Stack server.
     * Release stable version and support debugging. [v5.13.28](https://github.com/ossrs/srs-stack/releases/tag/v5.13.28)
     * HLS: Set m3u8 expire time to 1s for LLHLS. v5.13.29
     * Transcript: Support set the force_style for overlay subtitle. v5.13.30
+    * Transcript: Use Whisper response without LF. (#163). v5.13.31
 * v5.12
     * Refine local variable name conf to config. v5.12.1
     * Add forced exit on timeout for program termination. v5.12.1

diff --git a/platform/transcript.go b/platform/transcript.go
@@ -1646,26 +1646,8 @@ func (v *TranscriptTask) DriveAsrQueue(ctx context.Context) error {
 		srt.WriteString(fmt.Sprintf("%02d:%02d:%02d,%03d\n",
 			int(e.Hours()), int(e.Minutes())%60, int(e.Seconds())%60, int(e.Milliseconds())%1000))
 
-		// Limit each line of text, write a new line if exceed.
-		lineMaxSize := 45
-		words := strings.Split(srtSegment.Text, " ")
-		var current string
-		for _, word := range words {
-			if word == "" {
-				continue
-			}
-
-			if len(current)+len(word) < lineMaxSize {
-				current += word + " "
-				continue
-			}
-
-			srt.WriteString(fmt.Sprintf("%v\n", current))
-			current = word + " "
-		}
-		if current != "" {
-			srt.WriteString(fmt.Sprintf("%v\n", current))
-		}
+		// Write the subtitle text.
+		srt.WriteString(fmt.Sprintf("%v\n", srtSegment.Text))
 
 		// Insert a new line.
 		srt.WriteString("\n")

diff --git a/ui/src/pages/ScenarioTranscript.js b/ui/src/pages/ScenarioTranscript.js
@@ -192,6 +192,9 @@ function ScenarioTranscriptImpl({activeKey, defaultEnabled, defaultConf, default
             size: Number(segment.size / 1024.0),
             eac: Number(segment.eac),
             asrc: Number(segment.asrc),
+            // The max length of the subtitle text in asrs array.
+            asrsMaxLength: Math.max(...segment.asrs.map(asr => asr.text.length)),
+            asrsMaxWords: Math.max(...segment.asrs.map(asr => asr.text.split(' ').length)),
             // Rules:
             // 1. Always allow to clear the first segment, that is only one segment in the queue.
             // 2. Prevent the first segment from clearing subtitles, as it may have already been added
@@ -225,6 +228,9 @@ function ScenarioTranscriptImpl({activeKey, defaultEnabled, defaultConf, default
             eac: Number(segment.eac),
             asrc: Number(segment.asrc),
             olc: Number(segment.olc),
+            // The max length of the subtitle text in asrs array.
+            asrsMaxLength: Math.max(...segment.asrs.map(asr => asr.text.length)),
+            asrsMaxWords: Math.max(...segment.asrs.map(asr => asr.text.split(' ').length)),
           };
         });
         setOverlayQueue(queue);
@@ -420,6 +426,7 @@ function ScenarioTranscriptImpl({activeKey, defaultEnabled, defaultConf, default
                 <th>Duration</th>
                 <th title={t('transcript.eac')}>EAC</th>
                 <th title={t('transcript.asrc')}>ASRC</th>
+                <th>Segments</th>
                 <th>Size</th>
                 <th>Text</th>
                 <th>{t('transcript.action')}</th>
@@ -433,6 +440,9 @@ function ScenarioTranscriptImpl({activeKey, defaultEnabled, defaultConf, default
                   <td>{`${segment.duration.toFixed(1)}`}s</td>
                   <td>{`${segment.eac.toFixed(1)}`}ms</td>
                   <td>{`${segment.asrc.toFixed(1)}`}ms</td>
+                  <td title={`There are ${segment.asrs.length} segments, max text length is ${segment.asrsMaxLength} bytes, max words is ${segment.asrsMaxWords}`}>
+                    {segment.asrs.length}/{segment.asrsMaxLength}/{segment.asrsMaxWords}
+                  </td>
                   <td>{`${segment.size.toFixed(1)}`}KB</td>
                   <td style={{textDecoration: segment.uca ? "line-through" : ''}}>{segment.asr}</td>
                   <td>
@@ -464,6 +474,7 @@ function ScenarioTranscriptImpl({activeKey, defaultEnabled, defaultConf, default
                 <th title={t('transcript.eac')}>EAC</th>
                 <th title={t('transcript.asrc')}>ASRC</th>
                 <th title={t('transcript.olc')}>OLC</th>
+                <th>Segments</th>
                 <th>Size</th>
                 <th>Text</th>
               </tr>
@@ -478,6 +489,9 @@ function ScenarioTranscriptImpl({activeKey, defaultEnabled, defaultConf, default
                   <td>{`${segment.asrc.toFixed(1)}`}ms</td>
                   <td>{`${segment.olc.toFixed(1)}`}ms</td>
                   <td>{`${segment.size.toFixed(1)}`}MB</td>
+                  <td title={`There are ${segment.asrs.length} segments, max text length is ${segment.asrsMaxLength} bytes, max words is ${segment.asrsMaxWords}`}>
+                    {segment.asrs.length}/{segment.asrsMaxLength}/{segment.asrsMaxWords}
+                  </td>
                   <td style={{textDecoration: segment.uca ? "line-through" : ''}}>{segment.asr}</td>
                 </tr>;
               })}