Update app.py

app.py

@@ -18,9 +18,9 @@ from groq import Groq
 
 client = Groq(api_key=os.getenv("GROQ_API_KEY"))
 
-TRANSCRIPTION_MODEL = "whisper-large-v3"  # updated model
+TRANSCRIPTION_MODEL = "turbowhisper-large-v3"
 REWRITE_MODEL = "openai/gpt-oss-120b"
-CHUNK_SIZE = 600  # 10 min in seconds
+CHUNK_SIZE = 600  # 10 min in seconds for audio splitting
 
 # ============================================================================
 # UTILITY FUNCTIONS
@@ -99,23 +99,41 @@ def transcribe_audio_chunks(audio_chunks, language="English"):
     return transcript_text.strip()
 
 # ============================================================================
-# GRAMMAR & PUNCTUATION CORRECTION
+# GRAMMAR & PUNCTUATION CORRECTION (CHUNKED)
 # ============================================================================
 
-def rewrite_transcript(text, language="English"):
-    """Rewrite transcript using GPT-OSS-120B with correct grammar & punctuation"""
+def rewrite_transcript_chunked(text, language="English", chunk_size=4000):
+    """
+    Rewrite transcript in smaller chunks to avoid token limits
+    chunk_size is approx number of words per request (adjust as needed)
+    """
     try:
-        prompt = f"Correct grammar, punctuation, and make it readable. Language: {language}\n\n{text}"
-        response = client.chat.completions.create(
-            model=REWRITE_MODEL,
-            messages=[{"role":"user", "content": prompt}],
-            temperature=0
-        )
-        rewritten_text = response.choices[0].message.content
+        words = text.split()
+        rewritten_chunks = []
+
+        for i in range(0, len(words), chunk_size):
+            chunk_text = " ".join(words[i:i+chunk_size])
+            prompt = f"Correct grammar, punctuation, and make it readable. Language: {language}\n\n{chunk_text}"
+            response = client.chat.completions.create(
+                model=REWRITE_MODEL,
+                messages=[{"role": "user", "content": prompt}],
+                temperature=0
+            )
+            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:
         return f"[Error rewriting transcript: {str(e)}]"
 
+def correct_transcript(transcript_text, language):
+    lang = "English" if "English" in language else "Urdu"
+    if not transcript_text:
+        return "", "Please generate transcript first." if lang=="English" else "پہلے ٹرانسکرپٹ بنائیں۔"
+    rewritten_text = rewrite_transcript_chunked(transcript_text, language=lang)
+    return rewritten_text, "✅ Grammar & punctuation corrected / گرامر اور پنکچویشن درست کی گئی"
+
 # ============================================================================
 # PROCESS VIDEO
 # ============================================================================
@@ -132,13 +150,6 @@ def generate_transcript(video_file, language):
         error_msg = str(e)
         return "", error_msg if lang=="English" else "خرابی: " + error_msg
 
-def correct_transcript(transcript_text, language):
-    lang = "English" if "English" in language else "Urdu"
-    if not transcript_text:
-        return "", "Please generate transcript first." if lang=="English" else "پہلے ٹرانسکرپٹ بنائیں۔"
-    rewritten_text = rewrite_transcript(transcript_text, language=lang)
-    return rewritten_text, "✅ Grammar & punctuation corrected / گرامر اور پنکچویشن درست کی گئی"
-
 # ============================================================================
 # DOWNLOAD FUNCTION
 # ============================================================================