Update app.py

app.py

@@ -20,23 +20,39 @@ def format_time_ms(milliseconds):
     return f"{hrs:02}:{mins:02}:{secs:02},{ms:03}"
 
 def smart_text_split(text, words_per_line, lines_per_segment):
-    # First split by major punctuation with more granular control
+    # Define natural break patterns
+    end_sentence = r'[.!?]+'
+    mid_sentence = r'[,;:]+'
+    
+    # First split by major punctuation
     sentences = []
     current = ""
     
-    # Split text into meaningful chunks using various punctuation marks
-    for char in text:
-        current += char
-        if char in '.!?':  # Strong break
-            sentences.append(current.strip())
-            current = ""
-        elif char in ',;:':  # Soft break - only break if it makes a meaningful chunk
-            if len(current.split()) >= words_per_line:
-                sentences.append(current.strip())
-                current = ""
-    
-    if current:
-        sentences.append(current.strip())
+    # Clean the text and ensure proper spacing after punctuation
+    text = re.sub(r'([.!?,;:])\s*', r'\1 ', text).strip()
+    
+    # Split into initial chunks by strong punctuation
+    chunks = re.split(f'({end_sentence})', text)
+    temp_sentences = []
+    
+    for i in range(0, len(chunks)-1, 2):
+        if i+1 < len(chunks):
+            temp_sentences.append(chunks[i] + chunks[i+1])
+        else:
+            temp_sentences.append(chunks[i])
+    
+    # Further process each sentence
+    for sentence in temp_sentences:
+        # Split by mid-sentence punctuation if sentence is too long
+        if len(sentence.split()) > words_per_line * 2:
+            sub_chunks = re.split(f'({mid_sentence})', sentence)
+            for i in range(0, len(sub_chunks)-1, 2):
+                if i+1 < len(sub_chunks):
+                    sentences.append(sub_chunks[i] + sub_chunks[i+1])
+                else:
+                    sentences.append(sub_chunks[i])
+        else:
+            sentences.append(sentence)
     
     # Process sentences into lines and segments
     segments = []
@@ -45,38 +61,28 @@ def smart_text_split(text, words_per_line, lines_per_segment):
     
     for sentence in sentences:
         words = sentence.strip().split()
-        i = 0
         
-        while i < len(words):
-            # Look ahead to find the best breaking point
-            look_ahead = min(words_per_line, len(words) - i)
-            best_break = look_ahead
+        while words:
+            # Determine natural break point
+            break_point = min(words_per_line, len(words))
             
-            # Find natural breaking points
-            for j in range(look_ahead - 1, 0, -1):
-                if any(words[i + j - 1].endswith(p) for p in ',;:.!?') or \
-                   any(words[i + j].startswith(p) for p in '([{'):
-                    best_break = j
+            # Look for natural breaks
+            for i in range(break_point-1, 0, -1):
+                if any(words[i-1].endswith(p) for p in '.!?,;:') or \
+                   any(words[i].startswith(p) for p in '([{'):
+                    break_point = i
                     break
             
-            # Add words to current line
-            current_line.extend(words[i:i + best_break])
+            current_line = words[:break_point]
+            words = words[break_point:]
             
-            if len(current_line) >= words_per_line or i + best_break >= len(words):
-                # Add line to current segment
-                current_segment.append(' '.join(current_line))
-                current_line = []
-                
-                # Check if segment is complete
-                if len(current_segment) >= lines_per_segment:
-                    segments.append('\n'.join(current_segment))
-                    current_segment = []
+            current_segment.append(' '.join(current_line))
             
-            i += best_break
+            if len(current_segment) >= lines_per_segment:
+                segments.append('\n'.join(current_segment))
+                current_segment = []
     
     # Handle remaining content
-    if current_line:
-        current_segment.append(' '.join(current_line))
     if current_segment:
         segments.append('\n'.join(current_segment))
     
@@ -160,9 +166,15 @@ async def generate_accurate_srt(text, voice, rate, pitch, words_per_line, lines_
     return srt_path, audio_path
 
 async def process_text(text, pitch, rate, voice, words_per_line, lines_per_segment):
-    pitch_str = f"{pitch}Hz" if pitch != 0 else "0Hz"
-    # Fix: Always include + sign for zero and positive rates
-    rate_str = f"+{rate}%" if rate >= 0 else f"{rate}%"
+    # Set default pitch and rate strings that work well
+    pitch_str = "+0Hz"  # neutral pitch
+    rate_str = "+0%"    # neutral rate
+    
+    # Only modify if user has changed values
+    if pitch != 0:
+        pitch_str = f"{pitch:+d}Hz"
+    if rate != 0:
+        rate_str = f"{rate:+d}%"
     
     srt_path, audio_path = await generate_accurate_srt(
         text, 
@@ -223,11 +235,11 @@ app = gr.Interface(
     fn=process_text,
     inputs=[
         gr.Textbox(label="Enter Text", lines=10),
-        gr.Slider(label="Pitch Adjustment (Hz)", minimum=-20, maximum=20, value=0, step=1),
-        gr.Slider(label="Rate Adjustment (%)", minimum=-50, maximum=50, value=0, step=1),
+        gr.Slider(label="Pitch Adjustment (Hz)", minimum=-10, maximum=10, value=0, step=1),
+        gr.Slider(label="Rate Adjustment (%)", minimum=-25, maximum=25, value=0, step=1),
         gr.Dropdown(label="Select Voice", choices=list(voice_options.keys()), value="Jenny Female"),
-        gr.Slider(label="Words per Line", minimum=1, maximum=15, value=8, step=1),
-        gr.Slider(label="Lines per Segment", minimum=1, maximum=5, value=2, step=1)
+        gr.Slider(label="Words per Line", minimum=3, maximum=12, value=6, step=1),
+        gr.Slider(label="Lines per Segment", minimum=1, maximum=4, value=2, step=1)
     ],
     outputs=[
         gr.File(label="Download SRT"),
@@ -235,7 +247,7 @@ app = gr.Interface(
         gr.Audio(label="Preview Audio")
     ],
     title="Advanced TTS with Configurable SRT Generation",
-    description="Generate perfectly synchronized audio and subtitles with custom segmentation control."
+    description="Generate perfectly synchronized audio and subtitles with natural speech patterns."
 )
 
 app.launch()