Update app.py

app.py

@@ -3,11 +3,11 @@ import streamlit as st
 from core.gemini_handler import GeminiHandler
 from core.visual_engine import VisualEngine 
 from core.prompt_engineering import (
-    create_cinematic_treatment_prompt, # UPDATED
-    construct_dalle_prompt,           # UPDATED
-    create_narration_script_prompt_enhanced, # UPDATED
-    create_scene_regeneration_prompt,   # UPDATED
-    create_visual_regeneration_prompt   # UPDATED
+    create_cinematic_treatment_prompt, 
+    construct_dalle_prompt,           
+    create_narration_script_prompt_enhanced, 
+    create_scene_regeneration_prompt,   
+    create_visual_regeneration_prompt   
 )
 import os
 
@@ -19,11 +19,10 @@ def load_api_key(key_name_streamlit, key_name_env):
     key = None; secrets_available = hasattr(st, 'secrets')
     try:
         if secrets_available and key_name_streamlit in st.secrets: key = st.secrets[key_name_streamlit]
-    except Exception: pass # Silently ignore if st.secrets access fails
+    except Exception: pass 
     if not key and key_name_env in os.environ: key = os.environ[key_name_env]
     return key
 
-# Initialize API Keys and handlers once
 if 'keys_loaded' not in st.session_state:
     st.session_state.GEMINI_API_KEY = load_api_key("GEMINI_API_KEY", "GEMINI_API_KEY")
     st.session_state.OPENAI_API_KEY = load_api_key("OPENAI_API_KEY", "OPENAI_API_KEY")
@@ -39,7 +38,6 @@ if 'keys_loaded' not in st.session_state:
     st.session_state.visual_engine.set_pexels_api_key(st.session_state.PEXELS_API_KEY)
     st.session_state.keys_loaded = True
 
-# Initialize other session state variables
 for key, default_val in [
     ('story_treatment_scenes', []), ('scene_dalle_prompts', []), ('generated_visual_paths', []),
     ('video_path', None), ('character_definitions', {}), ('global_style_additions', ""),
@@ -57,7 +55,6 @@ def initialize_new_project():
     st.session_state.narration_script_display = ""
 
 def generate_visual_for_scene_core(scene_index, scene_data, version=1):
-    # scene_data here is one scene from story_treatment_scenes
     dalle_prompt = construct_dalle_prompt(
         scene_data, 
         st.session_state.character_definitions, 
@@ -65,15 +62,13 @@ def generate_visual_for_scene_core(scene_index, scene_data, version=1):
     )
     if not dalle_prompt: return False
     
-    # Ensure lists are long enough
     while len(st.session_state.scene_dalle_prompts) <= scene_index: st.session_state.scene_dalle_prompts.append("")
     while len(st.session_state.generated_visual_paths) <= scene_index: st.session_state.generated_visual_paths.append(None)
         
     st.session_state.scene_dalle_prompts[scene_index] = dalle_prompt
     
     filename = f"scene_{scene_data.get('scene_number', scene_index+1)}_visual_v{version}.png"
-    # Pass the full scene_data to visual_engine for Pexels query construction if needed
-    img_path = st.session_state.visual_engine.generate_image_visual(dalle_prompt, scene_data, filename)
+    img_path = st.session_state.visual_engine.generate_image_visual(dalle_prompt, scene_data, filename) 
     
     if img_path and os.path.exists(img_path):
         st.session_state.generated_visual_paths[scene_index] = img_path; return True
@@ -87,7 +82,7 @@ with st.sidebar:
     user_idea = st.text_area("Core Story Idea / Theme:", "A lone wanderer searches for a mythical oasis in a vast, post-apocalyptic desert, haunted by mirages and mechanical scavengers.", height=120, key="user_idea_main")
     genre = st.selectbox("Primary Genre:", ["Cyberpunk", "Sci-Fi", "Fantasy", "Noir", "Thriller", "Western", "Post-Apocalyptic"], index=6, key="genre_main")
     mood = st.selectbox("Overall Mood:", ["Hopeful yet Desperate", "Mysterious & Eerie", "Gritty & Tense", "Epic & Awe-Inspiring", "Melancholy & Reflective"], index=0, key="mood_main")
-    num_scenes = st.slider("Number of Key Scenes:", 1, 4, 2, key="num_scenes_main") # Max 3-4 for API cost/time
+    num_scenes = st.slider("Number of Key Scenes:", 1, 4, 2, key="num_scenes_main")
 
     creative_guidance_options = {"Standard": "standard", "More Artistic": "more_artistic", "Experimental Narrative": "experimental_narrative"}
     selected_creative_guidance = st.selectbox("AI Creative Guidance Level:", options=list(creative_guidance_options.keys()), key="creative_guidance_select")
@@ -100,7 +95,7 @@ with st.sidebar:
                 status.write("Phase 1: Gemini crafting cinematic treatment (scenes, style, camera, sound)... 📜")
                 treatment_prompt = create_cinematic_treatment_prompt(user_idea, genre, mood, num_scenes, creative_guidance_options[selected_creative_guidance])
                 try:
-                    st.session_state.story_treatment_scenes = st.session_state.gemini_handler.generate_story_breakdown(treatment_prompt) # Re-use for JSON list
+                    st.session_state.story_treatment_scenes = st.session_state.gemini_handler.generate_story_breakdown(treatment_prompt)
                     num_gen_scenes = len(st.session_state.story_treatment_scenes)
                     st.session_state.scene_dalle_prompts = [""] * num_gen_scenes
                     st.session_state.generated_visual_paths = [None] * num_gen_scenes
@@ -116,7 +111,7 @@ with st.sidebar:
                     
                     status.update(label="Visuals ready! Generating narration script...", state="running")
                     narration_prompt = create_narration_script_prompt_enhanced(st.session_state.story_treatment_scenes, mood, genre, st.session_state.get("selected_voice_style", "cinematic_trailer"))
-                    narr_script = st.session_state.gemini_handler.generate_image_prompt(narration_prompt) # generate_image_prompt is just text gen
+                    narr_script = st.session_state.gemini_handler.generate_image_prompt(narration_prompt) 
                     st.session_state.narration_script_display = narr_script
                     status.update(label="Narration script ready! Synthesizing voice...", state="running")
                     
@@ -130,9 +125,7 @@ with st.sidebar:
 
     st.markdown("---")
     st.markdown("### Fine-Tuning Options")
-    # Character Definitions Expander (same as previous)
     with st.expander("Define Characters", expanded=False):
-        # ... (UI for char_name_input, char_desc_input, add button - same as previous, ensure lowercase keys for definitions)
         char_name_input = st.text_input("Character Name", key="char_name_adv_input_ultra")
         char_desc_input = st.text_area("Detailed Visual Description", key="char_desc_adv_input_ultra", height=100, placeholder="e.g., Jax: rugged male astronaut, mid-40s, salt-and-pepper short hair, cybernetic left eye glowing faintly blue, wearing a patched-up crimson flight suit.")
         if st.button("Save Character", key="add_char_adv_btn_ultra"):
@@ -142,9 +135,7 @@ with st.sidebar:
             st.caption("Current Characters:")
             for k,v in st.session_state.character_definitions.items(): st.markdown(f"**{k.title()}:** _{v}_")
 
-    # Style Controls Expander (same as previous)
     with st.expander("Global Style Overrides", expanded=False):
-        # ... (UI for predefined_styles, custom_style_keywords, apply button - same as previous) ...
         predefined_styles = { "Default (Director's Choice)": "", "Hyper-Realistic Gritty Noir": "hyper-realistic gritty neo-noir, extreme detail, deep dynamic shadows, complex reflections on wet surfaces, cinematic film grain, desaturated palette with isolated vibrant neon accents (e.g. red, cyan), anamorphic lens distortion, atmospheric haze.", "Surreal Dreamscape Fantasy": "surreal dreamscape, epic fantasy elements, painterly with photorealistic details, impossible architecture, bioluminescent flora, otherworldly color palette (e.g., magenta skies, turquoise rivers), style of Roger Dean meets Zdzisław Beksiński.", "Vintage Analog Sci-Fi": "70s/80s analog sci-fi film aesthetic, tangible practical effects look, subtle light leaks, lens flares, warm filmic tones mixed with cool blues, detailed retro-futuristic technology with chunky buttons and CRT screens."}
         selected_preset = st.selectbox("Base Style Preset:", options=list(predefined_styles.keys()), key="style_preset_select_adv_ultra")
         custom_keywords = st.text_area("Additional Custom Style Keywords:", key="custom_style_keywords_adv_input_ultra", height=80, placeholder="e.g., 'shot with Arri Alexa, shallow depth of field, golden hour tones'")
@@ -157,10 +148,7 @@ with st.sidebar:
             else: st.info("Global style additions cleared (using Director's per-scene choice).")
         if current_style_desc: st.caption(f"Active global style additions: \"{current_style_desc}\"")
 
-    # Voice Customization Expander (same as previous)
     with st.expander("Voice Customization (ElevenLabs)", expanded=False):
-        # ... (UI for available_voices_conceptual, selected_voice, set button - same) ...
-        # In a real app, fetch from ElevenLabs API if key is present. This is a placeholder list.
         elevenlabs_voices = ["Rachel", "Adam", "Bella", "Antoni", "Elli", "Josh", "Arnold", "Domi", "Fin", "Sarah"] 
         current_el_voice = st.session_state.visual_engine.elevenlabs_voice_id if hasattr(st.session_state, 'visual_engine') else "Rachel"
         selected_el_voice = st.selectbox("Narrator Voice:", elevenlabs_voices, index=elevenlabs_voices.index(current_el_voice) if current_el_voice in elevenlabs_voices else 0, key="el_voice_select_ultra")
@@ -168,11 +156,9 @@ with st.sidebar:
             if hasattr(st.session_state, 'visual_engine'): st.session_state.visual_engine.elevenlabs_voice_id = selected_el_voice
             st.success(f"Narrator voice set to: {selected_el_voice}")
 
-
 # --- Main Content Area ---
 st.header("🎬 Cinematic Storyboard & Treatment")
 
-# Display Narration Script if available
 if st.session_state.narration_script_display:
     with st.expander("📜 View Full Narration Script", expanded=False):
         st.markdown(f"> _{st.session_state.narration_script_display}_")
@@ -183,15 +169,13 @@ else:
     for i_main, scene_content_display in enumerate(st.session_state.story_treatment_scenes):
         scene_num = scene_content_display.get('scene_number', i_main + 1)
         scene_title = scene_content_display.get('scene_title', 'Untitled Scene')
-        unique_key_base = f"scene_{scene_num}_{''.join(filter(str.isalnum, scene_title[:10]))}" # For unique widget keys
+        unique_key_base = f"scene_{scene_num}_{''.join(filter(str.isalnum, scene_title[:10]))}"
 
-        st.subheader(f"SCENE {scene_num}: {scene_title.upper()}")
-        
-        # Display Director's Notes if any (for experimental narrative)
         if "director_note" in scene_content_display:
             st.info(f"🎬 Director's Note for Scene {scene_num}: {scene_content_display['director_note']}")
 
-        col_details, col_visual = st.columns([0.45, 0.55]) # Adjust ratio
+        st.subheader(f"SCENE {scene_num}: {scene_title.upper()}")
+        col_details, col_visual = st.columns([0.45, 0.55])
 
         with col_details:
             with st.expander("📝 Scene Treatment Details", expanded=True):
@@ -205,25 +189,21 @@ else:
                 st.markdown(f"**🎬 Director's Visual Style:** {scene_content_display.get('PROACTIVE_visual_style_감독', 'N/A')}")
                 st.markdown(f"**🎥 Director's Camera Work:** {scene_content_display.get('PROACTIVE_camera_work_감독', 'N/A')}")
                 st.markdown(f"**🔊 Director's Sound Design:** {scene_content_display.get('PROACTIVE_sound_design_감독', 'N/A')}")
-                
                 current_dalle_prompt = st.session_state.scene_dalle_prompts[i_main] if i_main < len(st.session_state.scene_dalle_prompts) else None
                 if current_dalle_prompt:
                      with st.popover("👁️ View DALL-E Prompt"): 
                         st.markdown(f"**Full DALL-E Prompt:**"); st.code(current_dalle_prompt, language='text')
-                
                 pexels_query_display = scene_content_display.get('pexels_search_query_감독', None)
                 if pexels_query_display:
                     st.caption(f"Suggested Pexels Query: `{pexels_query_display}`")
-
-
+        
         with col_visual:
             current_img_path = st.session_state.generated_visual_paths[i_main] if i_main < len(st.session_state.generated_visual_paths) else None
             if current_img_path and os.path.exists(current_img_path):
                 st.image(current_img_path, caption=f"Visual Concept for Scene {scene_num}: {scene_title}", use_column_width='always')
             else:
-                st.caption("Visual concept pending or failed.")
+                if st.session_state.story_treatment_scenes: st.caption("Visual concept pending or failed.")
 
-            # Edit Popovers - logic for calling regeneration needs to use the new prompt structures
             with st.popover(f"✏️ Edit Scene {scene_num} Treatment"): 
                 feedback_script_edit = st.text_area("Describe changes to treatment details:", key=f"treat_feed_{unique_key_base}", height=150) 
                 if st.button(f"🔄 Update Scene {scene_num} Treatment", key=f"regen_treat_btn_{unique_key_base}"): 
@@ -231,13 +211,18 @@ else:
                         with st.status(f"Updating Scene {scene_num} Treatment...", expanded=True) as status_treat_regen:
                             regen_prompt_text = create_scene_regeneration_prompt(scene_content_display, feedback_script_edit, st.session_state.story_treatment_scenes)
                             try:
-                                updated_scene_data = st.session_state.gemini_handler.regenerate_scene_script_details(regen_prompt_text) # Assumes this returns one scene obj
+                                updated_scene_data = st.session_state.gemini_handler.regenerate_scene_script_details(regen_prompt_text)
                                 st.session_state.story_treatment_scenes[i_main] = updated_scene_data 
                                 status_treat_regen.update(label="Treatment updated! Regenerating visual & DALL-E prompt...", state="running")
+                                
                                 version_num = 1
-                                if current_img_path: try: base,_=os.path.splitext(os.path.basename(current_img_path)); version_num = int(base.split('_v')[-1])+1 if '_v' in base else 2 except: version_num=2
+                                if current_img_path:
+                                    try: 
+                                        base,_=os.path.splitext(os.path.basename(current_img_path))
+                                        if '_v' in base: version_num = int(base.split('_v')[-1])+1
+                                        else: version_num = 2 # If no _v, start with v2 for regenerated
+                                    except: version_num=2 # Fallback if parsing fails
                                 
-                                # Regenerate visual for the updated scene data
                                 if generate_visual_for_scene_core(i_main, updated_scene_data, version=version_num):
                                     status_treat_regen.update(label="Scene Treatment & Visual Updated! 🎉", state="complete", expanded=False)
                                 else: status_treat_regen.update(label="Treatment updated, visual failed.", state="warning", expanded=False)
@@ -252,20 +237,23 @@ else:
                 if st.button(f"🔄 Update Scene {scene_num} Visual Prompt & Image", key=f"regen_visual_btn_{unique_key_base}"): 
                     if feedback_visual_edit:
                         with st.status(f"Refining DALL-E prompt & regenerating visual...", expanded=True) as status_visual_edit_regen:
-                            # Gemini refines the DALL-E prompt
                             refinement_req_prompt = create_visual_regeneration_prompt(
                                 dalle_prompt_to_edit, feedback_visual_edit, scene_content_display, 
                                 st.session_state.character_definitions, st.session_state.global_style_additions
                             )
                             try:
-                                refined_dalle_prompt = st.session_state.gemini_handler.generate_image_prompt(refinement_req_prompt) # generate_image_prompt is just text gen
-                                st.session_state.scene_dalle_prompts[i_main] = refined_dalle_prompt # Update stored DALL-E prompt
+                                refined_dalle_prompt = st.session_state.gemini_handler.generate_image_prompt(refinement_req_prompt)
+                                st.session_state.scene_dalle_prompts[i_main] = refined_dalle_prompt 
                                 status_visual_edit_regen.update(label="DALL-E prompt refined! Regenerating visual...", state="running")
                                 version_num = 1
-                                if current_img_path: try: base,_=os.path.splitext(os.path.basename(current_img_path)); version_num = int(base.split('_v')[-1])+1 if '_v' in base else 2 except: version_num=2
+                                if current_img_path: 
+                                    try: 
+                                        base,_=os.path.splitext(os.path.basename(current_img_path))
+                                        if '_v' in base: version_num = int(base.split('_v')[-1])+1
+                                        else: version_num = 2
+                                    except: version_num=2
                                 
-                                # Generate visual with the NEW refined DALL-E prompt
-                                if generate_visual_for_scene_core(i_main, scene_content_display, version=version_num): # scene_content_display is still relevant context
+                                if generate_visual_for_scene_core(i_main, scene_content_display, version=version_num):
                                     status_visual_edit_regen.update(label="Visual Updated! 🎉", state="complete", expanded=False)
                                 else: status_visual_edit_regen.update(label="Prompt refined, visual failed.", state="warning", expanded=False)
                                 st.rerun()
@@ -273,7 +261,6 @@ else:
                     else: st.warning("Please provide feedback for visual prompt regeneration.")
         st.markdown("---")
     
-    # Video Generation Button
     if st.session_state.story_treatment_scenes and any(p for p in st.session_state.generated_visual_paths if p is not None):
         if st.button("🎬 Assemble Narrated Cinematic Animatic", key="assemble_ultra_video_btn", type="primary", use_container_width=True):
             with st.status("Assembling Ultra Animatic...", expanded=True) as status_vid:
@@ -283,7 +270,7 @@ else:
                     if img_p and os.path.exists(img_p):
                         image_data_for_vid.append({
                             'path':img_p, 'scene_num':scene_c.get('scene_number',i_vid+1), 
-                            'key_action':scene_c.get('key_plot_beat','') # Use key_plot_beat for overlay
+                            'key_action':scene_c.get('key_plot_beat','') 
                         }); status_vid.write(f"Adding Scene {scene_c.get('scene_number', i_vid + 1)} to video.")
                 
                 if image_data_for_vid:
@@ -292,7 +279,7 @@ else:
                         image_data_for_vid, 
                         overall_narration_path=st.session_state.overall_narration_audio_path,
                         output_filename="cinegen_ultra_animatic.mp4",
-                        duration_per_image=st.session_state.visual_engine.video_overlay_font_size * 0.15, # Dynamic duration based on text overlay font size
+                        duration_per_image=5, 
                         fps=24 
                     )
                     if st.session_state.video_path and os.path.exists(st.session_state.video_path):
@@ -301,7 +288,6 @@ else:
                 else: status_vid.update(label="No valid images for video.", state="error", expanded=False)
     elif st.session_state.story_treatment_scenes: st.info("Generate visuals before assembling video.")
 
-    # Video display and download
     if st.session_state.video_path and os.path.exists(st.session_state.video_path):
         st.header("🎬 Generated Cinematic Animatic")
         try: