Update app.py

app.py

@@ -8,10 +8,13 @@ import torch
 from datetime import timedelta
 from pyannote.audio import Pipeline
 
-# --- Configuration ---
-ENV_API_KEY = os.environ.get("GEMINI_API_KEY", "")
+# --- Configuration & Tokens ---
+# You can hardcode your token here or use the Sidebar in the App
+HF_TOKEN = os.environ.get("HF_TOKEN", "REPLACE_WITH_YOUR_HF_TOKEN")
+GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY", "")
 
 def format_timecode(seconds, fps=25):
+    """Converts seconds to HH:MM:SS:FF for Resolve/Premiere."""
     td = timedelta(seconds=seconds)
     total_seconds = int(td.total_seconds())
     hours = total_seconds // 3600
@@ -21,6 +24,7 @@ def format_timecode(seconds, fps=25):
     return f"{hours:02}:{minutes:02}:{secs:02}:{frames:02}"
 
 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
     for i, seg in enumerate(segments, 1):
@@ -29,6 +33,7 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25):
         duration = seg['src_end'] - seg['src_start']
         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"* FROM CLIP NAME: {source_name}")
         edl_lines.append(f"* {seg.get('note', 'Clip')}\n")
@@ -36,70 +41,86 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25):
     return "\n".join(edl_lines)
 
 def call_gemini_for_edl(transcript_data, story_prompt, api_key):
+    """Sends diarized, word-level transcript to Gemini Senior Editor."""
     if not api_key:
-        st.error("Gemini API Key is missing. Please provide it in the sidebar.")
+        st.error("Gemini API Key is missing.")
         return None
+        
     url = f"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-preview-09-2025:generateContent?key={api_key}"
+    
     system_prompt = (
-        "You are an expert Documentary Senior Editor. Use the provided transcript JSON (which includes speaker IDs and word-level timestamps) to create a condensed story. "
+        "You are an expert Documentary Senior Editor. Use the provided transcript JSON "
+        "(which includes Speaker IDs and word-level timestamps) to create a condensed story. "
         "Output ONLY a valid JSON array of segments with 'src_start', 'src_end', and 'note'. "
-        "Remove fluff, repeats, and interviewer interruptions. Focus on the most engaging content."
+        "Rules: Remove fluff/repeats, ignore interviewer interruptions, and focus on the hook."
     )
+    
     payload = {
-        "contents": [{"parts": [{"text": f"Brief: {story_prompt}\n\nTranscript:\n{json.dumps(transcript_data)}"}]}],
+        "contents": [{"parts": [{"text": f"Creative Brief: {story_prompt}\n\nTranscript Data:\n{json.dumps(transcript_data)}"}]}],
         "systemInstruction": {"parts": [{"text": system_prompt}]},
         "generationConfig": {"responseMimeType": "application/json"}
     }
+    
     try:
         res = requests.post(url, json=payload)
         res.raise_for_status()
         return json.loads(res.json()['candidates'][0]['content']['parts'][0]['text'])
     except Exception as e:
-        st.error(f"Error calling Gemini: {e}")
+        st.error(f"Senior Editor AI Error: {e}")
         return None
 
-# --- App Interface ---
+# --- Streamlit UI ---
 st.set_page_config(page_title="DocAI Editor", layout="wide")
-st.title("Documentary AI: Senior Editor Pipeline")
+st.title("Documentary AI: Pipeline")
 
 with st.sidebar:
     st.header("Settings")
-    user_key = st.text_input("Gemini API Key", type="password")
-    hf_token = st.text_input("Hugging Face Token (for Diarization)", type="password", help="Required for pyannote models.")
-    active_api_key = user_key if user_key else ENV_API_KEY
+    input_gemini_key = st.text_input("Gemini API Key", type="password", value=GEMINI_API_KEY)
+    input_hf_token = st.text_input("HF Token (Diarization)", type="password", value=HF_TOKEN)
+    
+    active_api_key = input_gemini_key if input_gemini_key else GEMINI_API_KEY
+    active_hf_token = input_hf_token if input_hf_token else HF_TOKEN
     fps = st.number_input("Timeline FPS", value=25)
 
-uploaded_file = st.file_uploader("Upload Video/Audio", type=["mp4", "m4a", "wav", "mp3", "mov"])
+uploaded_file = st.file_uploader("Upload Video/Audio Clip", type=["mp4", "m4a", "wav", "mp3", "mov"])
 
 if uploaded_file:
+    # --- Step 1: Technical Processing ---
     if "transcript" not in st.session_state:
         if st.button("Step 1: Transcribe & Diarize"):
-            if not hf_token:
-                st.error("Please provide a Hugging Face Token in the sidebar for diarization.")
+            if not active_hf_token or active_hf_token.startswith("REPLACE"):
+                st.error("Please provide a valid Hugging Face Token.")
             else:
-                with st.spinner("Processing... This includes audio extraction, speaker separation, and word-level transcription:"):
-                    # Save file
+                with st.spinner("Processing... Extracting audio, identifying speakers, and transcribing:"):
+                    # Save local temp file
                     with open("temp_input", "wb") as f:
                         f.write(uploaded_file.getbuffer())
                     
-                    # Extract audio
+                    # Optimized Audio: m4a, 64kbps, 16kHz, mono
                     subprocess.run([
-                        "ffmpeg", "-i", "temp_input", "-vn", "-acodec", "pcm_s16le", 
-                        "-ar", "16000", "-ac", "1", "audio.wav", "-y"
+                        "ffmpeg", "-i", "temp_input", 
+                        "-vn", "-acodec", "aac", "-ab", "64k", "-ar", "16000", "-ac", "1", 
+                        "audio_optimized.m4a", "-y"
                     ])
                     
                     # 1. Diarization
-                    pipeline = Pipeline.from_pretrained("pyannote/speaker-diarization@2.1", use_auth_token=hf_token)
-                    diarization = pipeline("audio.wav")
+                    try:
+                        pipeline = Pipeline.from_pretrained(
+                            "pyannote/speaker-diarization@2.1", 
+                            use_auth_token=active_hf_token
+                        )
+                        diarization = pipeline("audio_optimized.m4a")
+                    except Exception as e:
+                        st.error(f"Diarization Error: {e}")
+                        st.stop()
                     
-                    # 2. Whisper Transcription (Word Level)
+                    # 2. Whisper Transcription
                     model = whisper.load_model("base")
-                    result = model.transcribe("audio.wav", word_timestamps=True)
+                    result = model.transcribe("audio_optimized.m4a", word_timestamps=True)
                     
-                    # 3. Alignment Logic
+                    # 3. Alignment
                     final_segments = []
                     for segment in result['segments']:
-                        # Determine speaker for this segment based on mid-point
                         mid_time = (segment['start'] + segment['end']) / 2
                         speaker = "Unknown"
                         for turn, _, speaker_id in diarization.itertracks(yield_label=True):
@@ -112,7 +133,7 @@ if uploaded_file:
                             "text": segment['text'],
                             "start": segment['start'],
                             "end": segment['end'],
-                            "words": segment.get('words', []) # Word-level timestamps
+                            "words": segment.get('words', [])
                         })
                     
                     st.session_state.transcript = final_segments
@@ -120,16 +141,16 @@ if uploaded_file:
 
     if "transcript" in st.session_state:
         st.divider()
-        brief = st.text_area("Creative Brief", placeholder="e.g., Focus on the story about yeast, remove the interviewer.")
+        brief = st.text_area("Creative Brief", placeholder="e.g. Focus on the yeast story, remove the interviewer.")
         
         if st.button("Step 2: Create EDL"):
             if not active_api_key:
                 st.error("Gemini API Key required.")
             else:
-                with st.spinner("Senior Editor analyzing..."):
-                    segments = call_gemini_for_edl(st.session_state.transcript, brief, active_api_key)
-                    if segments:
-                        edl_content = generate_cmx_edl("AI_Edit", segments, uploaded_file.name, fps)
+                with st.spinner("Analyzing..."):
+                    edl_segments = call_gemini_for_edl(st.session_state.transcript, brief, active_api_key)
+                    if edl_segments:
+                        final_edl = generate_cmx_edl("AI_Senior_Editor_Cut", edl_segments, uploaded_file.name, fps)
                         st.subheader("EDL Preview")
-                        st.text_area("Output", value=edl_content, height=250)
-                        st.download_button("Download EDL", data=edl_content, file_name="edit.edl")
+                        st.code(final_edl, language="text")
+                        st.download_button("Download EDL", data=final_edl, file_name="edit.edl")