# app.py 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, construct_dalle_prompt, create_narration_script_prompt_enhanced, create_scene_regeneration_prompt, create_visual_regeneration_prompt ) import os import logging # --- Configuration & Initialization --- st.set_page_config(page_title="CineGen AI Ultra+", layout="wide", initial_sidebar_state="expanded") logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s') logger = logging.getLogger(__name__) # --- Global State Variables & API Key Setup --- def load_api_key(key_name_streamlit, key_name_env, service_name): key = None; secrets_available = hasattr(st, 'secrets') try: if secrets_available and key_name_streamlit in st.secrets: key = st.secrets[key_name_streamlit] if key: logger.info(f"{service_name} API Key loaded from Streamlit secrets.") except Exception as e: logger.warning(f"Could not access st.secrets for {key_name_streamlit}: {e}") if not key and key_name_env in os.environ: key = os.environ[key_name_env] if key: logger.info(f"{service_name} API Key loaded from environment variable.") if not key: logger.warning(f"{service_name} API Key NOT FOUND.") return key if 'services_initialized' not in st.session_state: logger.info("Initializing services and API keys...") st.session_state.GEMINI_API_KEY = load_api_key("GEMINI_API_KEY", "GEMINI_API_KEY", "Gemini") st.session_state.OPENAI_API_KEY = load_api_key("OPENAI_API_KEY", "OPENAI_API_KEY", "OpenAI/DALL-E") st.session_state.ELEVENLABS_API_KEY = load_api_key("ELEVENLABS_API_KEY", "ELEVENLABS_API_KEY", "ElevenLabs") st.session_state.PEXELS_API_KEY = load_api_key("PEXELS_API_KEY", "PEXELS_API_KEY", "Pexels") if not st.session_state.GEMINI_API_KEY: st.error("CRITICAL: Gemini API Key missing!"); st.stop() try: st.session_state.gemini_handler = GeminiHandler(api_key=st.session_state.GEMINI_API_KEY) logger.info("GeminiHandler initialized.") except Exception as e: st.error(f"Failed to init GeminiHandler: {e}"); logger.error(f"GeminiHandler init failed: {e}"); st.stop() try: st.session_state.visual_engine = VisualEngine(output_dir="temp_cinegen_media") st.session_state.visual_engine.set_openai_api_key(st.session_state.OPENAI_API_KEY) st.session_state.visual_engine.set_elevenlabs_api_key(st.session_state.ELEVENLABS_API_KEY) st.session_state.visual_engine.set_pexels_api_key(st.session_state.PEXELS_API_KEY) logger.info("VisualEngine initialized and API keys set.") except Exception as e: st.error(f"Failed to init VisualEngine or set keys: {e}"); logger.error(f"VisualEngine init/key setting failed: {e}") st.warning("VisualEngine issue. Features might use placeholders/be disabled.") st.session_state.services_initialized = True; logger.info("Service initialization complete.") for key, default_val in [ ('story_treatment_scenes', []), ('scene_dalle_prompts', []), ('generated_visual_paths', []), ('video_path', None), ('character_definitions', {}), ('global_style_additions', ""), ('overall_narration_audio_path', None), ('narration_script_display', "") ]: if key not in st.session_state: st.session_state[key] = default_val # --- Helper Functions --- (initialize_new_project, generate_visual_for_scene_core - same as previous) def initialize_new_project(): st.session_state.story_treatment_scenes, st.session_state.scene_dalle_prompts, st.session_state.generated_visual_paths = [], [], [] st.session_state.video_path, st.session_state.overall_narration_audio_path, st.session_state.narration_script_display = None, None, "" logger.info("New project initialized.") def generate_visual_for_scene_core(scene_index, scene_data, version=1): dalle_prompt = construct_dalle_prompt(scene_data, st.session_state.character_definitions, st.session_state.global_style_additions) if not dalle_prompt: logger.error(f"DALL-E prompt construction failed for scene {scene_data.get('scene_number', scene_index+1)}"); return False 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" 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; logger.info(f"Visual generated for Scene {scene_data.get('scene_number', scene_index+1)}: {img_path}"); return True else: st.session_state.generated_visual_paths[scene_index] = None; logger.warning(f"Visual generation FAILED for Scene {scene_data.get('scene_number', scene_index+1)}. img_path: {img_path}"); return False # --- UI Sidebar --- (same as previous full app.py) with st.sidebar: st.title("🎬 CineGen AI Ultra+"); st.markdown("### Creative Seed") 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", "Historical Drama", "Surreal"], index=6, key="genre_main") mood = st.selectbox("Overall Mood:", ["Hopeful yet Desperate", "Mysterious & Eerie", "Gritty & Tense", "Epic & Awe-Inspiring", "Melancholy & Reflective", "Whimsical & Lighthearted"], index=0, key="mood_main") num_scenes = st.slider("Number of Key Scenes:", 1, 3, 1, key="num_scenes_main") creative_guidance_options = {"Standard Director": "standard", "Artistic Visionary": "more_artistic", "Experimental Storyteller": "experimental_narrative"} selected_creative_guidance_key = st.selectbox("AI Creative Director Style:", options=list(creative_guidance_options.keys()), key="creative_guidance_select") actual_creative_guidance = creative_guidance_options[selected_creative_guidance_key] if st.button("🌌 Generate Cinematic Treatment", type="primary", key="generate_treatment_btn", use_container_width=True): initialize_new_project() if not user_idea.strip(): st.warning("Please provide a story idea.") else: with st.status("AI Director is envisioning your masterpiece...", expanded=True) as status: try: status.write("Phase 1: Gemini crafting cinematic treatment... 📜"); logger.info("Phase 1: Cinematic Treatment Gen.") treatment_prompt = create_cinematic_treatment_prompt(user_idea, genre, mood, num_scenes, actual_creative_guidance) treatment_result_json = st.session_state.gemini_handler.generate_story_breakdown(treatment_prompt) if not isinstance(treatment_result_json, list) or not treatment_result_json: raise ValueError("Gemini returned invalid scene list.") st.session_state.story_treatment_scenes = treatment_result_json; 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 logger.info(f"Phase 1 complete. {num_gen_scenes} scenes.") status.update(label="Treatment complete! ✅ Generating visuals...", state="running") status.write("Phase 2: Creating visuals (DALL-E/Pexels)... 🖼️"); logger.info("Phase 2: Visual Gen.") visual_successes = 0 for i, sc in enumerate(st.session_state.story_treatment_scenes): status.write(f" Visual for Scene {sc.get('scene_number', i+1)}: {sc.get('scene_title','Untitled')}..."); logger.info(f" Processing visual for Scene {sc.get('scene_number', i+1)}.") if generate_visual_for_scene_core(i, sc, version=1): visual_successes += 1 if visual_successes==0 and num_gen_scenes>0: logger.error("Visual gen failed all scenes."); status.update(label="Visual gen failed. Check logs & API.", state="error", expanded=True); st.stop() elif visual_successes < num_gen_scenes: logger.warning(f"Visuals partial ({visual_successes}/{num_gen_scenes})."); status.update(label=f"Visuals ready ({visual_successes}/{num_gen_scenes}). Generating narration...", state="running") else: logger.info("All visuals OK."); status.update(label="Visuals ready! Generating narration script...", state="running") status.write("Phase 3: Generating narration script... 🎤"); logger.info("Phase 3: Narration Script Gen.") voice_style = st.session_state.get("selected_voice_style_for_generation", "cinematic_trailer") narr_prompt = create_narration_script_prompt_enhanced(st.session_state.story_treatment_scenes, mood, genre, voice_style) st.session_state.narration_script_display = st.session_state.gemini_handler.generate_image_prompt(narr_prompt) logger.info("Narration script generated."); status.update(label="Narration script ready! Synthesizing voice...", state="running") status.write("Phase 4: Synthesizing voice (ElevenLabs)... 🔊"); logger.info("Phase 4: Voice Synthesis.") st.session_state.overall_narration_audio_path = st.session_state.visual_engine.generate_narration_audio(st.session_state.narration_script_display) if st.session_state.overall_narration_audio_path: logger.info("Voiceover OK."); status.update(label="Voiceover ready! ✨", state="running") else: logger.warning("Voiceover failed/skipped."); status.update(label="Voiceover failed/skipped.", state="warning") status.update(label="All components ready! View storyboard. 🚀", state="complete", expanded=False) except ValueError as ve: logger.error(f"ValueError: {ve}"); status.update(label=f"Input/Gemini response error: {ve}", state="error", expanded=True); except Exception as e: logger.error(f"Unhandled Exception: {e}", exc_info=True); status.update(label=f"Error: {e}", state="error", expanded=True); st.markdown("---"); st.markdown("### Fine-Tuning Options") # Advanced Options (same UI as previous) with st.expander("Define Characters", expanded=False): char_name = st.text_input("Character Name", key="char_name_adv_ultra"); char_desc = st.text_area("Visual Description", key="char_desc_adv_ultra", height=100, placeholder="e.g., Jax: rugged male astronaut...") if st.button("Save Character", key="add_char_adv_ultra"): if char_name and char_desc: st.session_state.character_definitions[char_name.strip().lower()] = char_desc.strip(); st.success(f"Char '{char_name.strip()}' saved.") else: st.warning("Name and description needed.") if st.session_state.character_definitions: st.caption("Current Characters:"); [st.markdown(f"**{k.title()}:** _{v}_") for k,v in st.session_state.character_definitions.items()] with st.expander("Global Style Overrides", expanded=False): presets = {"Default": "", "Gritty Noir": "gritty neo-noir...", "Epic Fantasy": "epic fantasy matte...", "Vintage Sci-Fi": "70s analog sci-fi..."} sel_preset = st.selectbox("Base Style Preset:", options=list(presets.keys()), key="style_preset_adv_ultra") custom_kw = st.text_area("Additional Custom Style Keywords:", key="custom_style_adv_ultra", height=80, placeholder="e.g., 'Dutch angle'") cur_style = st.session_state.global_style_additions if st.button("Apply Global Styles", key="apply_styles_adv_ultra"): final_s = presets[sel_preset]; if custom_kw.strip(): final_s = f"{final_s}, {custom_kw.strip()}" if final_s else custom_kw.strip() st.session_state.global_style_additions = final_s.strip(); cur_style = final_s.strip() if cur_style: st.success("Global styles applied!") else: st.info("Global style additions cleared.") if cur_style: st.caption(f"Active global styles: \"{cur_style}\"") with st.expander("Voice Customization (ElevenLabs)", expanded=False): el_voices = ["Rachel", "Adam", "Bella", "Antoni", "Elli", "Josh", "Arnold", "Domi"] engine_v_id = "Rachel"; if hasattr(st.session_state, 'visual_engine') and st.session_state.visual_engine: engine_v_id = st.session_state.visual_engine.elevenlabs_voice_id try: cur_v_idx = el_voices.index(engine_v_id) except ValueError: cur_v_idx = 0 sel_el_voice = st.selectbox("Narrator Voice:", el_voices, index=cur_v_idx, key="el_voice_sel_ultra") prompt_v_styles = {"Cinematic Trailer": "cinematic_trailer", "Neutral Documentary": "documentary_neutral", "Character Introspection": "introspective_character"} sel_prompt_v_style_key = st.selectbox("Narration Script Style:", list(prompt_v_styles.keys()), key="narr_style_sel") if st.button("Set Narrator Voice & Style", key="set_voice_btn_ultra"): if hasattr(st.session_state, 'visual_engine'): st.session_state.visual_engine.elevenlabs_voice_id = sel_el_voice st.session_state.selected_voice_style_for_generation = prompt_v_styles[sel_prompt_v_style_key] st.success(f"Narrator: {sel_el_voice}. Script Style: {sel_prompt_v_style_key}") # --- Main Content Area --- st.header("🎬 Cinematic Storyboard & Treatment") 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}_") if not st.session_state.story_treatment_scenes: st.info("Use sidebar to generate cinematic treatment.") else: for i, scene_content in enumerate(st.session_state.story_treatment_scenes): scene_n = scene_content.get('scene_number', i + 1); scene_t = scene_content.get('scene_title', 'Untitled') key_base = f"s{scene_n}_{''.join(filter(str.isalnum, scene_t[:10]))}" if "director_note" in scene_content and scene_content['director_note']: st.info(f"🎬 Director Note S{scene_n}: {scene_content['director_note']}") st.subheader(f"SCENE {scene_n}: {scene_t.upper()}"); col_d, col_v = st.columns([0.45, 0.55]) with col_d: with st.expander("📝 Scene Treatment", expanded=True): st.markdown(f"**Beat:** {scene_content.get('emotional_beat', 'N/A')}") st.markdown(f"**Setting:** {scene_content.get('setting_description', 'N/A')}") st.markdown(f"**Chars:** {', '.join(scene_content.get('characters_involved', ['N/A']))}") st.markdown(f"**Focus Moment:** _{scene_content.get('character_focus_moment', 'N/A')}_") st.markdown(f"**Plot Beat:** {scene_content.get('key_plot_beat', 'N/A')}") st.markdown(f"**Dialogue Hook:** `\"{scene_content.get('suggested_dialogue_hook', '...')}\"`") st.markdown("---"); st.markdown(f"**Dir. Visual Style:** _{scene_content.get('PROACTIVE_visual_style_감독', 'N/A')}_") st.markdown(f"**Dir. Camera:** _{scene_content.get('PROACTIVE_camera_work_감독', 'N/A')}_") st.markdown(f"**Dir. Sound:** _{scene_content.get('PROACTIVE_sound_design_감독', 'N/A')}_") cur_d_prompt = st.session_state.scene_dalle_prompts[i] if i < len(st.session_state.scene_dalle_prompts) else None if cur_d_prompt: with st.popover("👁️ DALL-E Prompt"): st.markdown(f"**DALL-E Prompt:**"); st.code(cur_d_prompt, language='text') pexels_q = scene_content.get('pexels_search_query_감독', None) if pexels_q: st.caption(f"Pexels Fallback Query: `{pexels_q}`") with col_v: cur_img_p = st.session_state.generated_visual_paths[i] if i < len(st.session_state.generated_visual_paths) else None if cur_img_p and os.path.exists(cur_img_p): st.image(cur_img_p, caption=f"Scene {scene_n}: {scene_t}", use_column_width='always') else: if st.session_state.story_treatment_scenes: st.caption("Visual pending/failed.") with st.popover(f"✏️ Edit Scene {scene_n} Treatment"): fb_script = st.text_area("Changes to treatment:", key=f"treat_fb_{key_base}", height=150) if st.button(f"🔄 Update Scene {scene_n} Treatment", key=f"regen_treat_btn_{key_base}"): if fb_script: with st.status(f"Updating Scene {scene_n}...", expanded=True) as s_treat_regen: prompt_text = create_scene_regeneration_prompt(scene_content, fb_script, st.session_state.story_treatment_scenes) try: updated_sc_data = st.session_state.gemini_handler.regenerate_scene_script_details(prompt_text) st.session_state.story_treatment_scenes[i] = updated_sc_data s_treat_regen.update(label="Treatment updated! Regenerating visual...", state="running") v_num = 1 if cur_img_p: try: b,_=os.path.splitext(os.path.basename(cur_img_p)); v_num = int(b.split('_v')[-1])+1 if '_v' in b else 2 except: v_num = 2 # Default to v2 if parsing fails else: v_num = 1 # First gen for this scene visual if generate_visual_for_scene_core(i, updated_sc_data, version=v_num): s_treat_regen.update(label="Treatment & Visual Updated! 🎉", state="complete", expanded=False) else: s_treat_regen.update(label="Treatment updated, visual failed.", state="warning", expanded=False) st.rerun() except Exception as e_regen: s_treat_regen.update(label=f"Error: {e_regen}", state="error") else: st.warning("Please provide feedback.") with st.popover(f"🎨 Edit Scene {scene_n} Visual Prompt"): d_prompt_edit = st.session_state.scene_dalle_prompts[i] if i < len(st.session_state.scene_dalle_prompts) else "No DALL-E prompt." st.caption("Current DALL-E Prompt:"); st.code(d_prompt_edit, language='text') fb_visual = st.text_area("Changes for DALL-E prompt:", key=f"visual_fb_{key_base}", height=150) if st.button(f"🔄 Update Scene {scene_n} Visual", key=f"regen_visual_btn_{key_base}"): if fb_visual: with st.status(f"Refining prompt & visual for Scene {scene_n}...", expanded=True) as s_visual_regen: ref_req_prompt = create_visual_regeneration_prompt(d_prompt_edit, fb_visual, scene_content, st.session_state.character_definitions, st.session_state.global_style_additions) try: refined_d_prompt = st.session_state.gemini_handler.generate_image_prompt(ref_req_prompt) st.session_state.scene_dalle_prompts[i] = refined_d_prompt s_visual_regen.update(label="DALL-E prompt refined! Regenerating visual...", state="running") v_num = 1 if cur_img_p: try: b,_=os.path.splitext(os.path.basename(cur_img_p)); v_num = int(b.split('_v')[-1])+1 if '_v' in b else 2 except: v_num=2 else: v_num = 1 if generate_visual_for_scene_core(i, scene_content, version=v_num): s_visual_regen.update(label="Visual Updated! 🎉", state="complete", expanded=False) else: s_visual_regen.update(label="Prompt refined, visual failed.", state="warning", expanded=False) st.rerun() except Exception as e_regen_vis: s_visual_regen.update(label=f"Error: {e_regen_vis}", state="error") else: st.warning("Please provide feedback.") st.markdown("---") 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: img_data_vid = [] for i_v, sc_c in enumerate(st.session_state.story_treatment_scenes): img_p_v = st.session_state.generated_visual_paths[i_v] if i_v < len(st.session_state.generated_visual_paths) else None if img_p_v and os.path.exists(img_p_v): img_data_vid.append({'path':img_p_v, 'scene_num':sc_c.get('scene_number',i_v+1), 'key_action':sc_c.get('key_plot_beat','')}); status_vid.write(f"Adding Scene {sc_c.get('scene_number', i_v+1)}.") if img_data_vid: status_vid.write("Calling video engine..."); st.session_state.video_path = st.session_state.visual_engine.create_video_from_images( img_data_vid, overall_narration_path=st.session_state.overall_narration_audio_path, output_filename="cinegen_ultra_animatic.mp4", duration_per_image=5, fps=24) if st.session_state.video_path and os.path.exists(st.session_state.video_path): status_vid.update(label="Ultra animatic assembled! 🎉", state="complete", expanded=False); st.balloons() else: status_vid.update(label="Video assembly failed. Check logs.", state="error", expanded=False) 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.") if st.session_state.video_path and os.path.exists(st.session_state.video_path): st.header("🎬 Generated Cinematic Animatic"); try: with open(st.session_state.video_path, 'rb') as vf_obj: video_bytes = vf_obj.read() st.video(video_bytes, format="video/mp4") with open(st.session_state.video_path, "rb") as fp_dl: st.download_button(label="Download Ultra Animatic", data=fp_dl, file_name=os.path.basename(st.session_state.video_path), mime="video/mp4", use_container_width=True, key="download_ultra_video_btn" ) except Exception as e: st.error(f"Error displaying video: {e}") # --- Footer --- st.sidebar.markdown("---"); st.sidebar.caption("CineGen AI Ultra+ | Visionary Cinematic Pre-Production")