Merge pull request #516 from nightcityblade/fix/issue-324

fix: improve subtitle readability

Audio-Transcription-Chrome/content.js

@@ -96,11 +96,11 @@ function init_element(lines = 3) {
 
     elem_container = document.createElement('div');
     elem_container.id = "transcription";
-    elem_container.style.cssText = 'padding-top:16px;font-size:18px;position: fixed; top: 85%; left: 50%; transform: translate(-50%, -50%);line-height:18px;width:500px;height:' + (captionLineCount * 30) + 'px;opacity:0.9;z-index:100;background:black;border-radius:10px;color:white;';
+    elem_container.style.cssText = 'padding:0 24px;font-family:Arial,Helvetica,sans-serif;font-size:22px;font-weight:600;line-height:30px;position:fixed;top:85%;left:50%;transform:translate(-50%,-50%);width:min(80vw,900px);min-height:' + (captionLineCount * 30) + 'px;z-index:2147483647;color:white;text-align:center;letter-spacing:0.01em;text-shadow:0 0 2px #000,0 2px 4px rgba(0,0,0,0.95);cursor:move;';
 
     for (var i = 0; i <= captionLineCount; i++) {
         elem_text = document.createElement('span');
-        elem_text.style.cssText = 'position: absolute;padding-left:16px;padding-right:16px;';
+        elem_text.style.cssText = 'position:absolute;left:50%;transform:translateX(-50%);max-width:100%;padding:2px 14px;background:rgba(0,0,0,0.72);border-radius:6px;box-decoration-break:clone;-webkit-box-decoration-break:clone;';
         elem_text.id = "t" + i;
         elem_container.appendChild(elem_text);
 

Audio-Transcription-Firefox/content.js

@@ -218,11 +218,11 @@ function init_element(lines = 3) {
 
     elem_container = document.createElement('div');
     elem_container.id = "transcription";
-    elem_container.style.cssText = 'padding-top:16px;font-size:18px;line-height:18px;position:fixed;top:85%;left:50%;transform:translate(-50%,-50%);width:500px;height:' + (captionLineCount * 30) + 'px;opacity:0.9;z-index:100;background:black;border-radius:10px;color:white;';
+    elem_container.style.cssText = 'padding:0 24px;font-family:Arial,Helvetica,sans-serif;font-size:22px;font-weight:600;line-height:30px;position:fixed;top:85%;left:50%;transform:translate(-50%,-50%);width:min(80vw,900px);min-height:' + (captionLineCount * 30) + 'px;z-index:2147483647;color:white;text-align:center;letter-spacing:0.01em;text-shadow:0 0 2px #000,0 2px 4px rgba(0,0,0,0.95);cursor:move;';
 
     for (var i = 0; i <= captionLineCount; i++) {
         elem_text = document.createElement('span');
-        elem_text.style.cssText = 'position: absolute;padding-left:16px;padding-right:16px;';
+        elem_text.style.cssText = 'position:absolute;left:50%;transform:translateX(-50%);max-width:100%;padding:2px 14px;background:rgba(0,0,0,0.72);border-radius:6px;box-decoration-break:clone;-webkit-box-decoration-break:clone;';
         elem_text.id = "t" + i;
         elem_container.appendChild(elem_text);
 

tests/test_client.py

@@ -4,9 +4,10 @@ import scipy
 import websocket
 import copy
 import unittest
+from io import StringIO
 from unittest.mock import patch, MagicMock
 from whisper_live.client import Client, TranscriptionClient, TranscriptionTeeClient
-from whisper_live.utils import resample
+from whisper_live.utils import print_transcript, resample
 from pathlib import Path
 
 
@@ -116,6 +117,33 @@ class TestAudioResampling(unittest.TestCase):
         os.remove(resampled_audio)
 
 
+class TestPrintTranscript(unittest.TestCase):
+    @patch("whisper_live.utils.shutil.get_terminal_size")
+    @patch("sys.stdout", new_callable=StringIO)
+    def test_print_transcript_respects_narrow_terminals(self, mock_stdout, mock_terminal_size):
+        mock_terminal_size.return_value = os.terminal_size((20, 20))
+
+        print_transcript(["This transcript should still wrap cleanly on a narrow terminal."])
+
+        output_lines = [line for line in mock_stdout.getvalue().splitlines() if line.strip()]
+        self.assertGreater(len(output_lines), 1)
+        self.assertTrue(all(len(line) <= 20 for line in output_lines))
+
+    @patch("whisper_live.utils.shutil.get_terminal_size")
+    @patch("sys.stdout", new_callable=StringIO)
+    def test_print_transcript_indents_timestamp_continuations(self, mock_stdout, mock_terminal_size):
+        mock_terminal_size.return_value = os.terminal_size((32, 20))
+
+        print_transcript(
+            [{"start": "00:00", "end": "00:05", "text": "This line should wrap and keep its timestamp indentation."}],
+            timestamps=True,
+        )
+
+        output_lines = [line.rstrip() for line in mock_stdout.getvalue().splitlines() if line.strip()]
+        self.assertGreater(len(output_lines), 1)
+        self.assertTrue(output_lines[1].startswith(" " * len("[00:00 -> 00:05] ")))
+
+
 class TestSendingAudioPacket(BaseTestCase):
     def test_send_packet(self):
         self.client.send_packet_to_server(self.mock_audio_packet)

whisper_live/utils.py

@@ -1,3 +1,5 @@
+import os
+import shutil
 import textwrap
 import scipy
 import numpy as np
@@ -11,15 +13,26 @@ def clear_screen():
 
 
 def print_transcript(text, translated=False, timestamps=False):
-    """Prints formatted transcript text."""
+    """Prints formatted transcript text in a subtitle-like block."""
+    terminal_width = shutil.get_terminal_size((80, 20)).columns
+    wrap_width = max(10, min(80, terminal_width - 8))
+
     if timestamps:
+        lines = []
         for t in text:
-            print(f'[{t["start"]} -> {t["end"]}] {t["text"]}')
+            prefix = f'[{t["start"]} -> {t["end"]}] '
+            wrapper = textwrap.TextWrapper(
+                width=wrap_width,
+                subsequent_indent=" " * len(prefix),
+            )
+            lines.extend(wrapper.wrap(f'{prefix}{t["text"]}'))
     else:
-        wrapper = textwrap.TextWrapper(width=60)
-        text=" ".join(text) if translated else "".join(text)
-        for line in wrapper.wrap(text=text):
-            print(line)
+        wrapper = textwrap.TextWrapper(width=wrap_width)
+        transcript = " ".join(text) if translated else "".join(text)
+        lines = wrapper.wrap(text=transcript)
+
+    for line in lines[-3:]:
+        print(line.center(terminal_width))
 
 
 def format_time(s):