Update app.py

app.py

@@ -106,6 +106,10 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25):
     """Constructs a CMX 3600 formatted EDL."""
     edl_lines = [f"TITLE: {edl_title}", "FCM: NON-DROP FRAME\n"]
     rec_start = 0.0
+    
+    # Sanitize source name for the Reel ID column
+    reel_id = source_name.replace(" ", "_")
+    
     for i, seg in enumerate(segments, 1):
         src_in = format_timecode(seg['src_start'], fps)
         src_out = format_timecode(seg['src_end'], fps)
@@ -113,7 +117,7 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25):
         rec_in = format_timecode(rec_start, fps)
         rec_out = format_timecode(rec_start + duration, fps)
         
-        edl_lines.append(f"{i:03}  AX       V     C        {src_in} {src_out} {rec_in} {rec_out}")
+        edl_lines.append(f"{i:03}  {reel_id}       V     C        {src_in} {src_out} {rec_in} {rec_out}")
         edl_lines.append(f"* FROM CLIP NAME: {source_name}")
         edl_lines.append(f"* {seg.get('note', 'Clip')}\n")
         rec_start += duration
@@ -155,13 +159,30 @@ def call_gemini_for_edl(transcript_data, story_prompt, api_key):
         return None
 
 # --- Streamlit UI ---
-st.set_page_config(page_title="DocAI Editor (WhisperX)", layout="wide")
-st.title("Documentary AI: Pipeline (WhisperX)")
+st.set_page_config(page_title="Junior Editor", layout="wide")
+st.title("Junior Editor")
+
+st.markdown("""
+**Instructions**
+* Upload your file here, either an video file or audio. 
+* Set your timeline FPS, choose the quality of your transcription and ,if you know the language, set this to speed up the transcription phase. 
+* Junior Editor will transcribe it and separate by speakers and then await your instruction. You can ask it to find the most engaging bits and put them together from “Speaker 1”, remove all of Speaker 2, or construct a narrative around whatever idea you choose. 
+* It will create an EDL to import back into your editing software. If you want it to reference a Master Clip rather than the proxy, type this in the side bar at the start. 
+""")
+
+st.divider()
 
 with st.sidebar:
     st.header("Project Settings")
     fps = st.number_input("Timeline FPS", value=25)
     
+    st.info("💡 **Conform Helper**")
+    custom_reel_name = st.text_input(
+        "EDL Reel Name", 
+        placeholder="Paste Raw File Name Here...", 
+        help="Leave empty to use the uploaded file name. Use this to link proxies to original camera files."
+    )
+
     st.header("Model Settings")
     model_size = st.selectbox("Whisper Model", ["large-v2", "medium", "base"], index=0)
     
@@ -188,103 +209,118 @@ with st.sidebar:
 uploaded_file = st.file_uploader("Upload Video/Audio Clip", type=["mp4", "m4a", "wav", "mp3", "mov"])
 
 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
+        st.session_state.last_processed_file = uploaded_file.name
+
+    # --- Auto-Process Logic ---
     if "transcript" not in st.session_state:
-        if st.button("Step 1: Transcribe & Diarize"):
-            if not ACTIVE_HF_TOKEN or "PASTE_YOUR_HF_TOKEN" in ACTIVE_HF_TOKEN:
-                st.error("Please provide a valid Hugging Face Token.")
-            else:
-                status_container = st.empty()
-                with status_container.container():
-                    st.write("🔄 **Processing Started...**")
-                    
+        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...**")
+                progress_bar = st.progress(0)
+                
+                try:
+                    # Save File
                     with open("temp_input", "wb") as f:
                         f.write(uploaded_file.getbuffer())
                     
-                    st.write("🎵 **Extracting Audio (WAV 16k Mono)...**")
+                    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"
                     ])
                     
-                    try:
-                        device = "cuda" if torch.cuda.is_available() else "cpu"
-                        if device == "cpu":
-                            st.warning("⚠️ No GPU detected. WhisperX will be very slow.")
-                        else:
-                            st.write(f"🚀 **Loading WhisperX on {device}...**")
-                        
-                        # 1. Transcribe
-                        compute_type = "float16" if device == "cuda" else "int8"
-                        model = whisperx.load_model(model_size, device, compute_type=compute_type)
-                        
-                        st.write("📝 **Transcribing...**")
-                        audio = whisperx.load_audio("temp_audio.wav")
-                        result = model.transcribe(audio, batch_size=16, language=target_language)
-                        
-                        # Memory cleanup
-                        del model
-                        gc.collect()
-                        torch.cuda.empty_cache()
-                        
-                        # 2. Align
-                        st.write("⏱️ **Aligning Audio...**")
-                        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()
-                        
-                        # 3. Diarize
-                        st.write("🗣️ **Diarizing Speakers...**")
-                        diarize_model = whisperx.DiarizationPipeline(use_auth_token=ACTIVE_HF_TOKEN, device=device)
-                        
-                        diarize_kwargs = {}
-                        if num_speakers > 0:
-                            diarize_kwargs = {"min_speakers": num_speakers, "max_speakers": num_speakers}
-                        
-                        diarize_segments = diarize_model(audio, **diarize_kwargs)
-                        
-                        # 4. Final Merge
-                        st.write("🔗 **Merging Transcripts...**")
-                        final_result = whisperx.assign_word_speakers(diarize_segments, result)
-                        
-                        processed_segments = []
-                        for segment in final_result["segments"]:
-                            processed_segments.append({
-                                "speaker": segment.get("speaker", "Unknown"),
-                                "text": segment["text"].strip(),
-                                "start": segment["start"],
-                                "end": segment["end"]
-                            })
-                        
-                        st.session_state.transcript = processed_segments
-                        st.success(f"Complete! Found {len(processed_segments)} segments.")
+                    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)
+                    
+                    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")
+                    result = model.transcribe(audio, batch_size=16, language=target_language)
+                    
+                    # Cleanup
+                    del model
+                    gc.collect()
+                    torch.cuda.empty_cache()
+                    
+                    # 2. Align
+                    st.write("⏱️ **Phase 3/4: Aligning Text...**")
+                    progress_bar.progress(60)
+                    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()
+                    
+                    # 3. Diarize
+                    st.write("🗣️ **Phase 4/4: Identifying Speakers...**")
+                    progress_bar.progress(80)
+                    diarize_model = whisperx.DiarizationPipeline(use_auth_token=ACTIVE_HF_TOKEN, device=device)
+                    
+                    diarize_kwargs = {}
+                    if num_speakers > 0:
+                        diarize_kwargs = {"min_speakers": num_speakers, "max_speakers": num_speakers}
+                    
+                    diarize_segments = diarize_model(audio, **diarize_kwargs)
+                    
+                    # 4. Final Merge
+                    st.write("🔗 **Finalizing...**")
+                    final_result = whisperx.assign_word_speakers(diarize_segments, result)
+                    
+                    processed_segments = []
+                    for segment in final_result["segments"]:
+                        processed_segments.append({
+                            "speaker": segment.get("speaker", "Unknown"),
+                            "text": segment["text"].strip(),
+                            "start": segment["start"],
+                            "end": segment["end"]
+                        })
+                    
+                    st.session_state.transcript = processed_segments
+                    progress_bar.progress(100)
+                    st.success(f"Done! Processed {len(processed_segments)} segments.")
 
-                    except Exception as e:
-                        st.error(f"Processing Error: {e}")
-                        # Clean up temp files
-                        if os.path.exists("temp_input"): os.remove("temp_input")
-                        if os.path.exists("temp_audio.wav"): os.remove("temp_audio.wav")
-                        st.stop()
+                except Exception as e:
+                    st.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 (Diarized)"):
-            for seg in st.session_state.transcript[:20]: 
+        with st.expander("Transcript Preview", expanded=True):
+            for seg in st.session_state.transcript: 
                 st.markdown(f"**{seg['speaker']}:** {seg['text']}")
         
-        brief = st.text_area("Creative Brief", placeholder="e.g. Focus on the yeast story.")
+        st.subheader("Your Instruction")
+        brief = st.text_area("What should the Junior Editor do?", placeholder="e.g. Find the most engaging bits and put them together from Speaker 1, or remove all of Speaker 2.")
         
-        if st.button("Step 2: Create EDL"):
+        if st.button("Generate Edit"):
             if not ACTIVE_GEMINI_KEY:
-                st.error("Gemini API Key required.")
+                st.error("Gemini API Key required in Secrets.")
             else:
-                with st.spinner("Analyzing..."):
+                with st.spinner("Junior Editor is thinking..."):
+                    final_source_name = custom_reel_name.strip() if custom_reel_name.strip() else uploaded_file.name
+                    
                     edl_segments = call_gemini_for_edl(st.session_state.transcript, brief, ACTIVE_GEMINI_KEY)
                     if edl_segments:
-                        final_edl = generate_cmx_edl("AI_Senior_Editor_Cut", edl_segments, uploaded_file.name, fps)
-                        st.subheader("EDL Preview")
+                        final_edl = generate_cmx_edl("Junior_Editor_Cut", edl_segments, final_source_name, fps)
+                        st.subheader("Ready for Import")
                         st.code(final_edl, language="text")
-                        st.download_button("Download EDL", data=final_edl, file_name="edit.edl")
+                        st.download_button("Download .EDL", data=final_edl, file_name="junior_editor_cut.edl")