Update app.py

app.py

@@ -1,5 +1,5 @@
-import re
 import numpy as np
+import re
 import concurrent.futures
 import gradio as gr
 from datetime import datetime
@@ -40,6 +40,15 @@ def silence(duration, fps=44100):
     Returns a silent AudioClip of the specified duration.
     """
     return AudioArrayClip(np.zeros((int(fps*duration), 2)), fps=fps)
+
+def count_words_or_characters(text):
+    # Count non-Chinese words
+    non_chinese_words = len(re.findall(r'\b[a-zA-Z0-9]+\b', text))
+    
+    # Count Chinese characters
+    chinese_chars = len(re.findall(r'[\u4e00-\u9fff]', text))
+    
+    return non_chinese_words + chinese_chars
     
 # Define the passcode
 PASSCODE = "show_feedback_db"
@@ -78,7 +87,6 @@ css = """
 }
 """
 
-
 # Function to save feedback or provide access to the database file
 def handle_feedback(feedback):
     feedback = feedback.strip()  # Clean up leading/trailing whitespace
@@ -123,15 +131,14 @@ def transcribe_video(video_path):
         start = segment["start"]
         end = segment["end"]
         text = segment["text"]
-        # Count words in the segment
-        word_count = len(re.findall(r'\w+', text))
+
         transcript_with_timestamps.append({
             "start": start,
             "end": end,
-            "text": text,
-            "word_count": word_count
+            "text": text
         })
-        
+
+        word_count = count_words_or_characters(text)
         total_words += word_count
         total_duration += (end - start)
     
@@ -270,13 +277,13 @@ def process_entry(entry, i, video_width, video_height, add_voiceover, target_lan
     audio_segment = None
     if add_voiceover:
         segment_audio_path = f"segment_{i}_voiceover.wav"
-        generate_voiceover_OpenAI([entry], target_language, segment_audio_path)
+        desired_duration = entry["end"] - entry["start"]
+        generate_voiceover_OpenAI([entry], target_language, desired_duration, segment_audio_path)
         audio_clip = AudioFileClip(segment_audio_path)
         # Get and log all methods in AudioFileClip        
         logger.info("Methods in AudioFileClip:")
         for method in dir(audio_clip):
             logger.info(method)
-        desired_duration = entry["end"] - entry["start"]
         
         # Log duration of the audio clip and the desired duration for debugging.
         logger.debug(f"Audio clip duration: {audio_clip.duration}, Desired duration: {desired_duration}")
@@ -355,7 +362,26 @@ def generate_voiceover(translated_json, language, output_audio_path):
     except Exception as e:
         raise ValueError(f"Error generating voiceover: {e}")
 
-def generate_voiceover_OpenAI(translated_json, language, output_audio_path):
+def truncated_linear(x):
+    if x < 15:
+        return 1
+    elif x > 25:
+        return 1.2
+    else:
+        slope = (1.2 - 1) / (25 - 15)
+        return 1 + slope * (x - 15)
+
+def calculate_speed(text, desired_duration):
+    # Calculate characters per second
+    char_count = len(text)
+    chars_per_second = char_count / (desired_duration + 0.001)
+    
+    # Apply truncated linear function to get speed
+    speed = truncated_linear(chars_per_second)
+    
+    return speed
+
+def generate_voiceover_OpenAI(translated_json, language, desired_duration, output_audio_path):
     """
     Generate voiceover from translated text for a given language using OpenAI TTS API.
     """
@@ -373,11 +399,13 @@ def generate_voiceover_OpenAI(translated_json, language, output_audio_path):
 
     while retry_count < max_retries:
         try:
+            speed_tts = calculate_speed(full_text, desired_duration)
             # Create the speech using OpenAI TTS API
             response = client.audio.speech.create(
                 model=model,
                 voice=voice,
-                input=full_text
+                input=full_text,
+                speed=speed_tts
             )
             # Save the audio to the specified path
             with open(output_audio_path, 'wb') as f: