Update app.py

app.py

@@ -4,6 +4,7 @@ import edge_tts
 import os
 import asyncio
 import uuid
+import re
 
 # Function to get the length of an audio file in seconds
 def get_audio_length(audio_file):
@@ -19,6 +20,32 @@ def format_time(seconds):
     secs = seconds % 60
     return f"{hrs:02}:{mins:02}:{secs:02},{millis:03}"
 
+# Function to split text into segments by punctuation or limit to 7-8 words
+def split_text_into_segments(text):
+    segments = []
+    # Split by punctuation (., !, ?)
+    raw_segments = re.split(r'([.!?])', text)
+    for i in range(0, len(raw_segments) - 1, 2):
+        # Combine segment with following punctuation
+        sentence = raw_segments[i].strip() + raw_segments[i + 1]
+        words = sentence.split()
+        
+        # If segment is longer than 8 words, split into 7-8 word chunks
+        if len(words) > 8:
+            for j in range(0, len(words), 8):
+                segments.append(" ".join(words[j:j+8]))
+        else:
+            segments.append(sentence.strip())
+    
+    # Handle remaining text after the last punctuation
+    if len(raw_segments) % 2 == 1:
+        remaining_text = raw_segments[-1].strip()
+        words = remaining_text.split()
+        for j in range(0, len(words), 8):
+            segments.append(" ".join(words[j:j+8]))
+    
+    return segments
+
 # Function to generate SRT with accurate timing per batch and cross-check timing
 async def generate_accurate_srt(batch_text, batch_num, start_offset):
     audio_file = f"batch_{batch_num}_audio.wav"
@@ -30,26 +57,23 @@ async def generate_accurate_srt(batch_text, batch_num, start_offset):
     # Get the actual length of the audio file
     actual_length = get_audio_length(audio_file)
 
-    # Initialize SRT content
-    srt_content = ""
-    words = batch_text.split()
-    num_segments = max(1, len(words) // 15)  # Group words into segments of ~15 words each
-    segment_duration = actual_length / num_segments  # Duration for each segment
+    # Split the text into segments based on punctuation and word count
+    segments = split_text_into_segments(batch_text)
+    segment_duration = actual_length / len(segments)  # Duration per segment
     start_time = start_offset
-    min_display_duration = 1.5  # Set a minimum display time of 1.5 seconds per subtitle
 
-    # Build SRT content with accurate timing, ensuring no segment exceeds audio length
-    for i in range(0, len(words), 15):
-        segment_words = words[i:i+15]
-        end_time = start_time + max(segment_duration, min_display_duration)
+    # Initialize SRT content
+    srt_content = ""
+    for index, segment in enumerate(segments):
+        end_time = start_time + segment_duration
         
         # If end_time exceeds actual audio length of the batch, adjust it
         if end_time > start_offset + actual_length:
             end_time = start_offset + actual_length
 
-        srt_content += f"{i // 15 + 1 + (batch_num * 100)}\n"
+        srt_content += f"{index + 1 + (batch_num * 100)}\n"
         srt_content += f"{format_time(start_time)} --> {format_time(end_time)}\n"
-        srt_content += " ".join(segment_words) + "\n\n"
+        srt_content += segment + "\n\n"
         
         # Update start time for next segment
         start_time = end_time