Update aduc_framework/engineers/deformes3D.py

aduc_framework/engineers/deformes3D.py

@@ -85,11 +85,27 @@ class Deformes3DEngine:
 
         logger.info(f"IMAGE SPECIALIST: Ordem para gerar {num_keyframes_to_generate} keyframes (versões LTX).")
 
+        ltx_conditioning_items0 = []
+
+        
+        img_pil0 = Image.open(initial_ref_path).convert("RGB")
+        img_processed0 = self._preprocess_image_for_latent_conversion(img_pil0, target_resolution_tuple)
+        pixel_tensor0 = self._pil_to_pixel_tensor(img_processed0)
+
+        ltx_conditioning_items0.append(LatentConditioningItem(pixel_tensor0, 0, 0.05))
+        ltx_conditioning_items0.append(LatentConditioningItem(pixel_tensor0, 24, 0.05))
+
+        latent_tensorY = pixel_tensor0
+        latent_tensorX = latent_tensorY
+
+        current_base_image_path = initial_ref_path
+        past_base_image_path = initial_ref_path
+        
+
         for i in range(num_keyframes_to_generate):
             scene_index = i + 1
             current_scene = storyboard[i]
             future_scene = storyboard[i + 1] if (i + 1) < len(storyboard) else "A cena final."
-            
             logger.info(f"--> Gerando Keyframe {scene_index}/{num_keyframes_to_generate}...")
             
             img_prompt = deformes2d_thinker_singleton.get_anticipatory_keyframe_prompt(
@@ -97,47 +113,48 @@ class Deformes3DEngine:
                 scene_history=previous_prompt,
                 current_scene_desc=current_scene,
                 future_scene_desc=future_scene,
-                last_image_path=current_base_image_path,
-                fixed_ref_paths=general_ref_paths
+                last_image_path=past_base_image_path,
+                fixed_ref_paths=current_base_image_path
             )
+
+            past_base_image_path = current_base_image_path
+
             
-            context_paths = [current_base_image_path] + [p for p in general_ref_paths if p != current_base_image_path][:3]
-            ltx_context_paths = list(reversed(context_paths))
-            
-            ltx_conditioning_items = []
-            weight = 0.6
-            for idx, path in enumerate(ltx_context_paths):
-                img_pil = Image.open(path).convert("RGB")
-                img_processed = self._preprocess_image_for_latent_conversion(img_pil, target_resolution_tuple)
-                pixel_tensor = self._pil_to_pixel_tensor(img_processed)
-                latent_tensor = vae_manager_singleton.encode(pixel_tensor)
-                ltx_conditioning_items.append(LatentConditioningItem(latent_tensor, 0, weight))
-                if idx >= 0: weight -= 0.1
+
+            ltx_conditioning_items = ltx_conditioning_items0
+            ltx_conditioning_items.append(LatentConditioningItem(latent_tensorX, 0, 0.4))
+            ltx_conditioning_items.append(LatentConditioningItem(latent_tensorY, 8, 0.6))
+
+            latent_tensorX = latent_tensorY
             
             ltx_base_params = {"guidance_scale": 1.0, "stg_scale": 0.001, "num_inference_steps": 25}
             generated_latents, _ = ltx_manager_singleton.generate_latent_fragment(
                 height=height, width=width,
                 conditioning_items_data=ltx_conditioning_items,
                 motion_prompt=img_prompt,
-                video_total_frames=48, video_fps=24,
+                video_total_frames=24, video_fps=24,
                 **ltx_base_params
             )
-
+            
             final_latent = generated_latents[:, :, -1:, :, :]
-            upscaled_latent = latent_enhancer_specialist_singleton.upscale(final_latent)
-            enriched_pixel_tensor = vae_manager_singleton.decode(upscaled_latent)
+            #upscaled_latent = latent_enhancer_specialist_singleton.upscale(final_latent)
+            enriched_pixel_tensor = vae_manager_singleton.decode(final_latent)
 
             pixel_path = os.path.join(self.workspace_dir, f"keyframe_{scene_index:04d}_pixel.png")
             latent_path = os.path.join(self.workspace_dir, f"keyframe_{scene_index:04d}_latent.pt")
             self.save_image_from_tensor(enriched_pixel_tensor, pixel_path)
             torch.save(final_latent.cpu(), latent_path)
 
+            latent_tensorY = latent_path
+
+            
             keyframe_data = {
                 "id": scene_index,
                 "caminho_pixel": pixel_path,
                 "caminho_latent": latent_path,
                 "prompt_keyframe": img_prompt
             }
+            
             all_keyframes_data.append(keyframe_data)
             
             if progress_callback: