Update app.py

app.py

@@ -163,7 +163,7 @@ async def generate_audio_with_voice_prefix(text_segment, default_voice, rate, pi
 
 async def process_transcript_line(line, default_voice, rate, pitch):
     """Processes a single transcript line with HH:MM:SS.milliseconds timestamp and quoted text segments."""
-    match = re.match(r'(\d{2}):(\d{2}):(\d{2})\.(\d{3})\s+(.*)', line)
+    match = re.match(r'(\d{2}):(\d{2}):(\d{2}),(\d{3})\s+(.*)', line) # Modified timestamp regex
     if match:
         hours, minutes, seconds, milliseconds, text_parts = match.groups()
         start_time_ms = (
@@ -201,8 +201,9 @@ async def transcript_to_speech(transcript_text, voice, rate, pitch):
     lines = transcript_text.strip().split('\n')
     timed_audio_segments = []
     max_end_time_ms = 0
+    previous_end_time_ms = 0
 
-    for line in lines:
+    for i, line in enumerate(lines):
         start_time, audio_paths = await process_transcript_line(line, voice, rate, pitch)
         if start_time is not None and audio_paths:
             combined_line_audio = AudioSegment.empty()
@@ -215,8 +216,27 @@ async def transcript_to_speech(transcript_text, voice, rate, pitch):
                     print(f"Warning: Audio file not found: {path}")
 
             if combined_line_audio:
-                timed_audio_segments.append({'start': start_time, 'audio': combined_line_audio})
-                max_end_time_ms = max(max_end_time_ms, start_time + len(combined_line_audio))
+                current_audio_duration = len(combined_line_audio)
+                intended_start_time = start_time
+
+                if i > 0:
+                    previous_line_match = re.match(r'(\d{2}):(\d{2}):(\d{2}),(\d{3})\s+.*', lines[i-1])
+                    if previous_line_match:
+                        prev_h, prev_m, prev_s, prev_ms, _ = previous_line_match.groups()
+                        previous_start_time_ms = (
+                            int(prev_h) * 3600000 +
+                            int(prev_m) * 60000 +
+                            int(prev_s) * 1000 +
+                            int(prev_ms)
+                        )
+                        time_difference = start_time - previous_start_time_ms
+
+                        if current_audio_duration > time_difference and timed_audio_segments:
+                            intended_start_time = previous_end_time_ms # Append to the previous audio
+
+                timed_audio_segments.append({'start': intended_start_time, 'audio': combined_line_audio})
+                previous_end_time_ms = intended_start_time + current_audio_duration
+                max_end_time_ms = max(max_end_time_ms, previous_end_time_ms)
         elif audio_paths:
             for path in audio_paths:
                 try:
@@ -244,12 +264,12 @@ async def create_demo():
     voices = await get_voices()
     default_voice = "en-US-AndrewMultilingualNeural - en-US (Male)"
     description = """
-    Process timestamped text (HH:MM:SS.milliseconds) with voice changes within quotes.
-    Format: `HH:MM:SS.milliseconds "VoicePrefix Text" more text "AnotherVoicePrefix More Text"`
+    Process timestamped text (HH:MM:SS,milliseconds) with voice changes within quotes.
+    Format: `HH:MM:SS,milliseconds "VoicePrefix Text" more text "AnotherVoicePrefix More Text"`
     Example:
     ```
-    00:00:00.000 "This is the default voice." more default. "1F Now a female voice." and back to default.
-    00:00:05.000 "1C Yes," said the child, "it is fun!"
+    00:00:00,000 "This is the default voice." more default. "1F Now a female voice." and back to default.
+    00:00:05,000 "1C Yes," said the child, "it is fun!"
     ```
     ***************************************************************************************************
     1M = en-AU-WilliamNeural - en-AU (Male)
@@ -271,16 +291,16 @@ async def create_demo():
     demo = gr.Interface(
         fn=tts_interface,
         inputs=[
-            gr.Textbox(label="Timestamped Text with Voice Changes", lines=10, placeholder='00:00:00.000 "Text" more text "1F Different Voice"'),
+            gr.Textbox(label="Timestamped Text with Voice Changes", lines=10, placeholder='00:00:00,000 "Text" more text "1F Different Voice"'),
             gr.Dropdown(choices=[""] + list(voices.keys()), label="Select Default Voice", value=default_voice),
             gr.Slider(minimum=-50, maximum=50, value=0, label="Speech Rate Adjustment (%)", step=1),
-            gr.Slider(minimum=-50, maximum=50, value=0, label="Pitch Adjustment (Hz)", step=1)
+            gr.Slider(minimum=-50, maximum=50, value=50, value=0, label="Pitch Adjustment (Hz)", step=1)
         ],
         outputs=[
             gr.Audio(label="Generated Audio", type="filepath"),
             gr.Markdown(label="Warning", visible=False)
         ],
-        title="TTS with HH:MM:SS.milliseconds and In-Quote Voice Switching",
+        title="TTS with HH:MM:SS,milliseconds and In-Quote Voice Switching",
         description=description,
         analytics_enabled=False,
         allow_flagging=False