Update core/visual_engine.py

core/visual_engine.py

@@ -1,5 +1,5 @@
 # core/visual_engine.py
-from PIL import Image, ImageDraw, ImageFont, ImageOps 
+from PIL import Image, ImageDraw, ImageFont, ImageOps
 from moviepy.editor import (ImageClip, concatenate_videoclips, TextClip, 
                             CompositeVideoClip, AudioFileClip)
 import moviepy.video.fx.all as vfx 
@@ -16,12 +16,26 @@ import logging
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.INFO)
 
-# --- ElevenLabs Import ---
+# --- MONKEY PATCH FOR Image.ANTIALIAS (LAST RESORT) ---
+# Uncomment this section ONLY if updating MoviePy/Pillow and removing Ken Burns vfx.resize doesn't work.
+# This is a hack to make old MoviePy code that calls Image.ANTIALIAS work with newer Pillow.
+# if hasattr(Image, 'Resampling') and hasattr(Image.Resampling, 'LANCZOS'): # Pillow 9+
+#     if not hasattr(Image, 'ANTIALIAS'):
+#         logger.warning("PIL.Image.ANTIALIAS not found. Monkey-patching with Image.Resampling.LANCZOS.")
+#         Image.ANTIALIAS = Image.Resampling.LANCZOS
+# elif hasattr(Image, 'LANCZOS'): # Pillow 8 used Image.LANCZOS directly
+#      if not hasattr(Image, 'ANTIALIAS'):
+#         logger.warning("PIL.Image.ANTIALIAS not found. Monkey-patching with Image.LANCZOS.")
+#         Image.ANTIALIAS = Image.LANCZOS
+# else:
+#     logger.warning("Pillow version is too old or Resampling attributes not found for ANTIALIAS monkey-patch.")
+# --- END MONKEY PATCH ---
+
+
 ELEVENLABS_CLIENT_IMPORTED = False
 ElevenLabsAPIClient = None 
 Voice = None              
 VoiceSettings = None      
-
 try:
     from elevenlabs.client import ElevenLabs as ImportedElevenLabsClient 
     from elevenlabs import Voice as ImportedVoice, VoiceSettings as ImportedVoiceSettings
@@ -29,53 +43,31 @@ try:
     Voice = ImportedVoice
     VoiceSettings = ImportedVoiceSettings
     ELEVENLABS_CLIENT_IMPORTED = True
-    logger.info("Successfully imported ElevenLabs client components (SDK v1.x.x pattern).")
-except ImportError as e_eleven:
-    logger.warning(f"Could not import ElevenLabs client components: {e_eleven}. ElevenLabs audio will be disabled.")
-except Exception as e_gen_eleven: 
-    logger.warning(f"General error importing ElevenLabs: {e_gen_eleven}. ElevenLabs audio will be disabled.")
-
+    logger.info("Successfully imported ElevenLabs client components.")
+except ImportError as e_eleven: logger.warning(f"ElevenLabs client import failed: {e_eleven}. Audio disabled.")
+except Exception as e_gen_eleven: logger.warning(f"General ElevenLabs import error: {e_gen_eleven}. Audio disabled.")
 
 class VisualEngine:
     def __init__(self, output_dir="temp_cinegen_media"):
-        self.output_dir = output_dir
-        os.makedirs(self.output_dir, exist_ok=True)
+        self.output_dir = output_dir; os.makedirs(self.output_dir, exist_ok=True)
+        self.font_filename = "arial.ttf"; self.font_path_in_container = f"/usr/local/share/fonts/truetype/mycustomfonts/{self.font_filename}"
+        self.font_size_pil = 20; self.video_overlay_font_size = 30; self.video_overlay_font_color = 'white'; self.video_overlay_font = 'Liberation-Sans-Bold'
+        try: self.font = ImageFont.truetype(self.font_path_in_container, self.font_size_pil); logger.info(f"Placeholder font: {self.font_path_in_container}.")
+        except IOError: logger.warning(f"Placeholder font '{self.font_path_in_container}' fail. Default."); self.font = ImageFont.load_default(); self.font_size_pil = 10
         
-        self.font_filename = "arial.ttf" 
-        self.font_path_in_container = f"/usr/local/share/fonts/truetype/mycustomfonts/{self.font_filename}"
-        self.font_size_pil = 20 
-        self.video_overlay_font_size = 30 
-        self.video_overlay_font_color = 'white'
-        self.video_overlay_font = 'Liberation-Sans-Bold' # More likely to be found by ImageMagick on Linux
-
-        try:
-            self.font = ImageFont.truetype(self.font_path_in_container, self.font_size_pil)
-            logger.info(f"Placeholder font loaded: {self.font_path_in_container}.")
-        except IOError:
-            logger.warning(f"Placeholder font '{self.font_path_in_container}' not found. Using default.")
-            self.font = ImageFont.load_default()
-            self.font_size_pil = 10
-
         self.openai_api_key = None; self.USE_AI_IMAGE_GENERATION = False
-        self.dalle_model = "dall-e-3"; self.image_size_dalle3 = "1792x1024" 
-        self.video_frame_size = (1280, 720) 
-
-        self.elevenlabs_api_key = None; self.USE_ELEVENLABS = False
-        self.elevenlabs_client = None 
-        self.elevenlabs_voice_id = "Rachel" 
+        self.dalle_model = "dall-e-3"; self.image_size_dalle3 = "1792x1024"; self.video_frame_size = (1280, 720)
+        self.elevenlabs_api_key = None; self.USE_ELEVENLABS = False; self.elevenlabs_client = None
+        # IMPORTANT: Replace with a VALID Voice ID from your ElevenLabs account for the model you use.
+        # Common pre-made voices might work by name, but IDs are more reliable.
+        self.elevenlabs_voice_id = "Rachel" # E.g., "21m00Tcm4TlvDq8ikWAM" or another valid ID
         if VoiceSettings and ELEVENLABS_CLIENT_IMPORTED: 
-            self.elevenlabs_voice_settings = VoiceSettings(
-                stability=0.60, similarity_boost=0.80, 
-                style=0.15, use_speaker_boost=True
-            )
+            self.elevenlabs_voice_settings = VoiceSettings(stability=0.60, similarity_boost=0.80, style=0.15, use_speaker_boost=True)
         else: self.elevenlabs_voice_settings = None
         self.pexels_api_key = None; self.USE_PEXELS = False
         logger.info("VisualEngine initialized.")
 
-    def set_openai_api_key(self,k): 
-        self.openai_api_key=k; self.USE_AI_IMAGE_GENERATION=bool(k)
-        logger.info(f"DALL-E ({self.dalle_model}) {'Ready.' if k else 'Disabled (no API key).'}")
-
+    def set_openai_api_key(self,k): self.openai_api_key=k; self.USE_AI_IMAGE_GENERATION=bool(k); logger.info(f"DALL-E {self.dalle_model} {'Ready.' if k else 'Disabled.'}")
     def set_elevenlabs_api_key(self,api_key):
         self.elevenlabs_api_key=api_key
         if api_key and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient: 
@@ -83,19 +75,12 @@ class VisualEngine:
                 self.elevenlabs_client = ElevenLabsAPIClient(api_key=api_key) 
                 if self.elevenlabs_client: self.USE_ELEVENLABS=True; logger.info("ElevenLabs Client Ready.")
                 else: self.USE_ELEVENLABS=False; logger.warning("ElevenLabs client is None post-init.")
-            except Exception as e: 
-                logger.error(f"Error initializing ElevenLabs client: {e}. Disabled.", exc_info=True); 
-                self.USE_ELEVENLABS=False; self.elevenlabs_client = None
-        else: 
-            self.USE_ELEVENLABS=False; self.elevenlabs_client = None
-            if not (ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient): pass # Logged at import
-            else: logger.info("ElevenLabs API Key not provided. Disabled.")
-            
-    def set_pexels_api_key(self,k):
-        self.pexels_api_key=k; self.USE_PEXELS=bool(k)
-        logger.info(f"Pexels Search {'Ready.' if k else 'Disabled (no API key).'}")
+            except Exception as e: logger.error(f"Error init ElevenLabs client: {e}. Disabled.", exc_info=True); self.USE_ELEVENLABS=False; self.elevenlabs_client = None
+        else: self.USE_ELEVENLABS=False; self.elevenlabs_client = None
+              
+    def set_pexels_api_key(self,k): self.pexels_api_key=k; self.USE_PEXELS=bool(k); logger.info(f"Pexels Search {'Ready.' if k else 'Disabled.'}")
     
-    def _get_text_dimensions(self,text_content,font_obj): # No changes
+    def _get_text_dimensions(self,text_content,font_obj): # Same
         if not text_content: return 0,self.font_size_pil 
         try:
             if hasattr(font_obj,'getbbox'): bbox=font_obj.getbbox(text_content);w=bbox[2]-bbox[0];h=bbox[3]-bbox[1];return w, h if h > 0 else self.font_size_pil
@@ -103,7 +88,7 @@ class VisualEngine:
             else: return int(len(text_content)*self.font_size_pil*0.6),int(self.font_size_pil*1.2 if self.font_size_pil*1.2>0 else self.font_size_pil)
         except: return int(len(text_content)*self.font_size_pil*0.6),int(self.font_size_pil*1.2)
     
-    def _create_placeholder_image_content(self,text_description,filename,size=None): # No changes
+    def _create_placeholder_image_content(self,text_description,filename,size=None): # Same
         if size is None: size = self.video_frame_size
         img=Image.new('RGB',size,color=(20,20,40));d=ImageDraw.Draw(img);padding=25;max_w=size[0]-(2*padding);lines=[];
         if not text_description: text_description="(Placeholder: No prompt text)"
@@ -126,7 +111,7 @@ class VisualEngine:
         try:img.save(filepath);return filepath
         except Exception as e:logger.error(f"Saving placeholder image {filepath}: {e}", exc_info=True);return None
 
-    def _search_pexels_image(self, query, output_filename_base): # No changes
+    def _search_pexels_image(self, query, output_filename_base): # Same
         if not self.USE_PEXELS or not self.pexels_api_key: return None
         headers = {"Authorization": self.pexels_api_key}; params = {"query": query, "per_page": 1, "orientation": "landscape", "size": "large"}
         pexels_filename = output_filename_base.replace(".png", f"_pexels_{random.randint(1000,9999)}.jpg")
@@ -145,7 +130,7 @@ class VisualEngine:
         except Exception as e: logger.error(f"Pexels search/download for query '{query}': {e}", exc_info=True)
         return None
 
-    def generate_image_visual(self, image_prompt_text, scene_data, scene_identifier_filename): # No changes
+    def generate_image_visual(self, image_prompt_text, scene_data, scene_identifier_filename): # Same
         filepath = os.path.join(self.output_dir, scene_identifier_filename)
         if self.USE_AI_IMAGE_GENERATION and self.openai_api_key:
             max_retries = 2
@@ -183,44 +168,34 @@ class VisualEngine:
         
         audio_filepath = os.path.join(self.output_dir, output_filename)
         try:
-            logger.info(f"Generating ElevenLabs audio (Voice: {self.elevenlabs_voice_id}) for: {text_to_narrate[:70]}...")
+            logger.info(f"Generating ElevenLabs audio (Voice ID: {self.elevenlabs_voice_id}) for: {text_to_narrate[:70]}...")
             
-            # Determine how to pass voice (as object or ID string)
-            voice_param = self.elevenlabs_voice_id 
-            if Voice and self.elevenlabs_voice_settings: # Check if Voice & VoiceSettings were imported
-                voice_param = Voice(
-                    voice_id=self.elevenlabs_voice_id, # This voice_id must be a valid ID string or a known name
-                    settings=self.elevenlabs_voice_settings
-                )
-            
-            # Use the text_to_speech.stream() method for newer SDK
+            # Using client.text_to_speech.stream()
             if hasattr(self.elevenlabs_client, 'text_to_speech') and hasattr(self.elevenlabs_client.text_to_speech, 'stream'):
                 logger.info("Using elevenlabs_client.text_to_speech.stream()")
+                # For text_to_speech.stream, voice_id must be a string. VoiceSettings are passed to client or model.
                 audio_data_iterator = self.elevenlabs_client.text_to_speech.stream(
                     text=text_to_narrate,
-                    voice_id=self.elevenlabs_voice_id, # stream usually takes voice_id string
-                    model_id="eleven_multilingual_v2" # or other models like "eleven_monolingual_v1"
-                )
-            # Fallback to direct .generate() if text_to_speech.stream isn't there (might be an older v1 client or different structure)
-            elif hasattr(self.elevenlabs_client, 'generate'):
-                logger.info("Using elevenlabs_client.generate() as fallback.")
-                audio_data_iterator = self.elevenlabs_client.generate(
-                    text=text_to_narrate,
-                    voice=voice_param, # This might take the Voice object or just the ID string
-                    model="eleven_multilingual_v2"
+                    voice_id=str(self.elevenlabs_voice_id), # Ensure it's a string
+                    model_id="eleven_multilingual_v2", # Or "eleven_monolingual_v1"
+                    # voice_settings=self.elevenlabs_voice_settings # Pass settings if API supports it here
                 )
+            elif hasattr(self.elevenlabs_client, 'generate') and Voice and self.elevenlabs_voice_settings:
+                 logger.info("Using elevenlabs_client.generate() with Voice object.")
+                 voice_param = Voice(voice_id=self.elevenlabs_voice_id, settings=self.elevenlabs_voice_settings)
+                 audio_data_iterator = self.elevenlabs_client.generate(
+                    text=text_to_narrate, voice=voice_param, model="eleven_multilingual_v2")
             else:
-                logger.error("No recognized audio generation method (text_to_speech.stream or generate) found on ElevenLabs client.")
+                logger.error("No recognized audio generation method found on ElevenLabs client. Update logic or check SDK.")
                 return None
             
             with open(audio_filepath, "wb") as f:
                 for chunk in audio_data_iterator: 
                     if chunk: f.write(chunk)
-            
             logger.info(f"ElevenLabs audio saved: {audio_filepath}")
             return audio_filepath
         except AttributeError as ae:
-             logger.error(f"AttributeError with ElevenLabs client: {ae}. Method or attribute likely missing for installed SDK version.", exc_info=True)
+             logger.error(f"AttributeError with ElevenLabs client: {ae}. SDK method/structure might be different.", exc_info=True)
         except Exception as e:
             logger.error(f"Error generating ElevenLabs audio: {e}", exc_info=True)
         return None
@@ -228,39 +203,39 @@ class VisualEngine:
     def create_video_from_images(self, image_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24, duration_per_image=4.5):
         if not image_data_list: logger.warning("No image data for video."); return None
         processed_clips = []; narration_audio_clip = None; final_video_clip_obj = None
+        logger.info(f"Preparing {len(image_data_list)} clips. Target frame: {self.video_frame_size}. Duration/img: {duration_per_image}s.")
 
-        logger.info(f"Preparing {len(image_data_list)} clips for video. Target frame size: {self.video_frame_size}")
         for i, data in enumerate(image_data_list):
             img_path, scene_num, key_action = data.get('path'), data.get('scene_num', i+1), data.get('key_action', '')
-            if not (img_path and os.path.exists(img_path)): logger.warning(f"Img not found or invalid: {img_path}"); continue
+            if not (img_path and os.path.exists(img_path)): logger.warning(f"Img not found: {img_path}"); continue
             try:
                 pil_img = Image.open(img_path)
                 if pil_img.mode != 'RGB': pil_img = pil_img.convert('RGB')
                 
-                # Resize and letterbox/pillarbox
                 img_copy = pil_img.copy()
-                # Use Image.Resampling.LANCZOS for high quality downscaling
+                # Use modern Image.Resampling.LANCZOS
                 img_copy.thumbnail(self.video_frame_size, Image.Resampling.LANCZOS) 
                 
                 canvas = Image.new('RGB', self.video_frame_size, (random.randint(0,5), random.randint(0,5), random.randint(0,5))) 
                 xo, yo = (self.video_frame_size[0]-img_copy.width)//2, (self.video_frame_size[1]-img_copy.height)//2
                 canvas.paste(img_copy, (xo,yo))
-                frame_np = np.array(canvas) # This numpy array is correctly sized.
+                frame_np = np.array(canvas)
                 
-                # ImageClip should not need to do much resizing now.
-                img_clip = ImageClip(frame_np).set_duration(duration_per_image)
+                img_clip_base = ImageClip(frame_np).set_duration(duration_per_image)
+
+                # TEST: Remove Ken Burns effect to isolate ANTIALIAS error source
+                # end_scale = random.uniform(1.03, 1.08) 
+                # img_clip = img_clip_base.fx(vfx.resize, lambda t: 1 + (end_scale - 1) * (t / duration_per_image))
+                # img_clip = img_clip.set_position('center')
+                img_clip = img_clip_base # Use base clip without resize effect for now
 
-                end_scale = random.uniform(1.03, 1.08) 
-                img_clip = img_clip.fx(vfx.resize, lambda t: 1 + (end_scale - 1) * (t / duration_per_image))
-                img_clip = img_clip.set_position('center')
-                
                 if key_action:
                     txt_clip = TextClip(f"Scene {scene_num}\n{key_action}", fontsize=self.video_overlay_font_size, 
                                         color=self.video_overlay_font_color, font=self.video_overlay_font,
                                         bg_color='rgba(10,10,20,0.8)', method='caption', align='West',
                                         size=(self.video_frame_size[0]*0.9, None), kerning=-1, stroke_color='black', stroke_width=1.5
                                        ).set_duration(duration_per_image - 1.0).set_start(0.5).set_position(('center', 0.92), relative=True)
-                    final_scene_clip = CompositeVideoClip([img_clip, txt_clip], size=self.video_frame_size, use_bgclip=True, bg_color=(0,0,0)) # Ensure bg for composite
+                    final_scene_clip = CompositeVideoClip([img_clip, txt_clip], size=self.video_frame_size, use_bgclip=True, bg_color=(0,0,0))
                 else: final_scene_clip = img_clip
                 processed_clips.append(final_scene_clip)
             except Exception as e: logger.error(f"Creating video clip for {img_path}: {e}", exc_info=True)
@@ -290,7 +265,7 @@ class VisualEngine:
                                             remove_temp=True, threads=os.cpu_count() or 2, logger='bar', bitrate="5000k")
             logger.info(f"Video successfully created: {output_path}"); return output_path
         except Exception as e: logger.error(f"Writing video file: {e}", exc_info=True); return None
-        finally: # Ensure clips are closed
+        finally:
             for c_item in processed_clips: 
                 if hasattr(c_item, 'close'): c_item.close()
             if narration_audio_clip and hasattr(narration_audio_clip, 'close'): narration_audio_clip.close()