Update generator.py

generator.py

@@ -810,7 +810,7 @@ class RetroArtConverter:
         # 1. InstantID (Identity)
         if self.instantid_active:
             if has_detected_faces and face_kps_image is not None:
-                # ... (code to append control_images is unchanged) ...
+                # Ensure face keypoints image has correct size
                 face_kps_image = ensure_correct_size(face_kps_image, target_width, target_height, "InstantID")
                 control_images.append(face_kps_image)
                 conditioning_scales.append(identity_control_scale)
@@ -830,16 +830,12 @@ class RetroArtConverter:
                         
                         print(f"  - Face embedding: {face_proj_embeds.shape}, Scale: {boosted_scale:.2f}")
                         
-                        # --- START FIX 2: Your padding solution ---
+                        # --- START FIX: Your padding solution ---
                         if 'prompt_embeds' in pipe_kwargs:
                             original_embeds = pipe_kwargs['prompt_embeds']
                             
-                            # Handle CFG by creating a [2, 16, 2048] tensor
-                            # [0] is zeros for negative, [1] is face embeds for positive
-                            face_proj_embeds_cfg = torch.cat([torch.zeros_like(face_proj_embeds), face_proj_embeds], dim=0)
-                            
                             # Concatenate face embeddings to POSITIVE prompt
-                            combined_embeds = torch.cat([original_embeds, face_proj_embeds_cfg[1:2]], dim=1) # [1, 93, 2048] -> [1, 109, 2048]
+                            combined_embeds = torch.cat([original_embeds, face_proj_embeds], dim=1)
                             pipe_kwargs['prompt_embeds'] = combined_embeds
 
                             # CRITICAL: Pad negative_prompt_embeds by the same amount
@@ -862,13 +858,13 @@ class RetroArtConverter:
                             print(f"  [OK] Face embeddings concatenated successfully! Prompt: {combined_embeds.shape}")
                         else:
                             print(f"  [WARNING] Can't concatenate - no prompt_embeds (use Compel)")
-                        # --- END FIX 2 ---
+                        # --- END FIX ---
                 
                 elif has_detected_faces:
                     print("  Face detected but IP-Adapter/embeddings unavailable, using keypoints only")
 
             else:
-                # ... (code for no face detected is unchanged) ...
+                # No face detected - blank map needed to maintain ControlNet list order
                 print("[INSTANTID] Using blank map (scale=0, no effect on generation)")
                 control_images.append(Image.new("RGB", (target_width, target_height), (0,0,0)))
                 conditioning_scales.append(0.0) # Set scale to 0