Update app.py

app.py

@@ -79,7 +79,6 @@ except Exception as e:
 if not hasattr(np, 'NaN'):
     np.NaN = np.nan
 
-# Import whisperx AFTER patching
 import whisperx
 
 # --- Configuration & Tokens ---
@@ -210,7 +209,6 @@ uploaded_file = st.file_uploader("Upload Video/Audio Clip", type=["mp4", "m4a",
 
 if uploaded_file:
     # --- Auto-Reset Logic for New Files ---
-    # If the user uploads a new file, we must clear the old transcript from memory
     if "last_processed_file" not in st.session_state or st.session_state.last_processed_file != uploaded_file.name:
         if "transcript" in st.session_state:
             del st.session_state.transcript
@@ -221,55 +219,62 @@ if uploaded_file:
         if not ACTIVE_HF_TOKEN or "PASTE_YOUR_HF_TOKEN" in ACTIVE_HF_TOKEN:
             st.error("Please provide a valid Hugging Face Token in the Sidebar/Secrets.")
         else:
-            status_container = st.empty()
-            with status_container.container():
-                st.info("🤖 **Junior Editor is starting...**")
+            # Create a dedicated container for the progress UI so it stays at the top
+            progress_container = st.container()
+            with progress_container:
+                st.info("🤖 **Junior Editor is processing your file...**")
+                
+                # Create the visual elements
+                status_text = st.empty()
                 progress_bar = st.progress(0)
                 
                 try:
-                    # Save File
+                    # Phase 1: Save File & Extract Audio
+                    status_text.markdown("**Phase 1/4: Extracting Audio...**")
                     with open("temp_input", "wb") as f:
                         f.write(uploaded_file.getbuffer())
                     
-                    st.write("🎵 **Phase 1/4: Extracting Audio...**")
-                    progress_bar.progress(10)
                     subprocess.run([
                         "ffmpeg", "-i", "temp_input", 
                         "-vn", "-acodec", "pcm_s16le", "-ar", "16000", "-ac", "1", 
                         "temp_audio.wav", "-y"
                     ])
+                    progress_bar.progress(25)
                     
+                    # Setup Device
                     device = "cuda" if torch.cuda.is_available() else "cpu"
                     if device == "cpu":
                         st.warning("⚠️ No GPU detected. This will be slow.")
                     
-                    # 1. Transcribe
-                    st.write(f"📝 **Phase 2/4: Transcribing (Whisper {model_size})...**")
-                    progress_bar.progress(30)
+                    # Phase 2: Transcribe
+                    status_text.markdown(f"**Phase 2/4: Transcribing (Whisper {model_size})... This is the longest step.**")
                     
                     compute_type = "float16" if device == "cuda" else "int8"
                     model = whisperx.load_model(model_size, device, compute_type=compute_type)
                     audio = whisperx.load_audio("temp_audio.wav")
+                    
+                    # This step takes the most time. It will sit at 25% until complete.
+                    # This is safer than chunking manually which can cause transcription errors.
                     result = model.transcribe(audio, batch_size=16, language=target_language)
                     
-                    # Cleanup
+                    # Cleanup VRAM
                     del model
                     gc.collect()
                     torch.cuda.empty_cache()
+                    progress_bar.progress(50)
                     
-                    # 2. Align
-                    st.write("⏱️ **Phase 3/4: Aligning Text...**")
-                    progress_bar.progress(60)
+                    # Phase 3: Align
+                    status_text.markdown("**Phase 3/4: Aligning Text...**")
                     model_a, metadata = whisperx.load_align_model(language_code=result["language"], device=device)
                     result = whisperx.align(result["segments"], model_a, metadata, audio, device, return_char_alignments=False)
                     
                     del model_a
                     gc.collect()
                     torch.cuda.empty_cache()
+                    progress_bar.progress(75)
                     
-                    # 3. Diarize
-                    st.write("🗣️ **Phase 4/4: Identifying Speakers...**")
-                    progress_bar.progress(80)
+                    # Phase 4: Diarize
+                    status_text.markdown("**Phase 4/4: Identifying Speakers...**")
                     diarize_model = whisperx.DiarizationPipeline(use_auth_token=ACTIVE_HF_TOKEN, device=device)
                     
                     diarize_kwargs = {}
@@ -278,8 +283,8 @@ if uploaded_file:
                     
                     diarize_segments = diarize_model(audio, **diarize_kwargs)
                     
-                    # 4. Final Merge
-                    st.write("🔗 **Finalizing...**")
+                    # Final Merge
+                    status_text.markdown("**Finalizing...**")
                     final_result = whisperx.assign_word_speakers(diarize_segments, result)
                     
                     processed_segments = []
@@ -292,16 +297,20 @@ if uploaded_file:
                         })
                     
                     st.session_state.transcript = processed_segments
+                    
+                    # --- CLEANUP ON SUCCESS ---
+                    if os.path.exists("temp_input"): os.remove("temp_input")
+                    if os.path.exists("temp_audio.wav"): os.remove("temp_audio.wav")
+                    
                     progress_bar.progress(100)
-                    st.success(f"Done! Processed {len(processed_segments)} segments.")
+                    status_text.success(f"Done! Processed {len(processed_segments)} segments.")
 
                 except Exception as e:
-                    st.error(f"Error during processing: {e}")
+                    status_text.error(f"Error during processing: {e}")
                     if os.path.exists("temp_input"): os.remove("temp_input")
                     if os.path.exists("temp_audio.wav"): os.remove("temp_audio.wav")
                     st.stop()
 
-    # --- Display Results if Transcript Exists ---
     if "transcript" in st.session_state:
         st.divider()
         with st.expander("Transcript Preview", expanded=True):