Update app.py

app.py

@@ -18,16 +18,16 @@ from groq import Groq
 
 client = Groq(api_key=os.getenv("GROQ_API_KEY"))
 
-TRANSCRIPTION_MODEL = "turbowhisper-large-v3"
+TRANSCRIPTION_MODEL = "whisper-large-v3"  # transcription model
 REWRITE_MODEL = "openai/gpt-oss-120b"
 CHUNK_SIZE = 600  # 10 min in seconds for audio splitting
+REWRITE_CHUNK_WORDS = 4000  # number of words per GPT rewrite chunk
 
 # ============================================================================
 # UTILITY FUNCTIONS
 # ============================================================================
 
 def get_video_duration(video_path):
-    """Get video duration using ffprobe"""
     try:
         result = subprocess.run(
             ["ffprobe", "-v", "error", "-show_entries",
@@ -43,20 +43,17 @@ def get_video_duration(video_path):
         raise RuntimeError(f"Failed to get video duration: {str(e)}")
 
 def extract_audio_chunks(video_path, chunk_size=CHUNK_SIZE):
-    """Extract audio from video and split into chunks"""
     try:
         duration = get_video_duration(video_path)
         num_chunks = math.ceil(duration / chunk_size)
         audio_chunks = []
 
-        # extract full audio first
         temp_audio_full = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
         audio_full_path = temp_audio_full.name
         temp_audio_full.close()
 
         os.system(f"ffmpeg -y -i \"{video_path}\" -q:a 0 -map a \"{audio_full_path}\"")
 
-        # split audio into chunks
         for i in range(num_chunks):
             start = i * chunk_size
             end = min((i + 1) * chunk_size, duration)
@@ -66,7 +63,6 @@ def extract_audio_chunks(video_path, chunk_size=CHUNK_SIZE):
             os.system(f"ffmpeg -y -i \"{audio_full_path}\" -ss {start} -to {end} -c copy \"{chunk_path}\"")
             audio_chunks.append(chunk_path)
 
-        # remove full audio file
         if os.path.exists(audio_full_path):
             os.unlink(audio_full_path)
 
@@ -102,17 +98,16 @@ def transcribe_audio_chunks(audio_chunks, language="English"):
 # GRAMMAR & PUNCTUATION CORRECTION (CHUNKED)
 # ============================================================================
 
-def rewrite_transcript_chunked(text, language="English", chunk_size=4000):
+def rewrite_transcript_chunked(text, language="English", chunk_words=REWRITE_CHUNK_WORDS):
     """
-    Rewrite transcript in smaller chunks to avoid token limits
-    chunk_size is approx number of words per request (adjust as needed)
+    Rewrite transcript in smaller word-based chunks to avoid token limits
     """
     try:
         words = text.split()
         rewritten_chunks = []
 
-        for i in range(0, len(words), chunk_size):
-            chunk_text = " ".join(words[i:i+chunk_size])
+        for i in range(0, len(words), chunk_words):
+            chunk_text = " ".join(words[i:i+chunk_words])
             prompt = f"Correct grammar, punctuation, and make it readable. Language: {language}\n\n{chunk_text}"
             response = client.chat.completions.create(
                 model=REWRITE_MODEL,
@@ -121,7 +116,6 @@ def rewrite_transcript_chunked(text, language="English", chunk_size=4000):
             )
             rewritten_chunks.append(response.choices[0].message.content)
 
-        # Join all rewritten chunks
         rewritten_text = "\n\n".join(rewritten_chunks)
         return rewritten_text
     except Exception as e: