Update generator.py

generator.py

@@ -458,10 +458,18 @@ class RetroArtConverter:
             # Configure ControlNets + IP-Adapter (SIMPLIFIED!)
             # First, check if we need to disable IP-Adapter
             if not (has_detected_faces and face_embeddings is not None):
-                # No faces or embeddings - disable IP-Adapter completely
+                # No faces - create dummy embeddings to prevent pipeline error
+                # The pipeline expects embeddings even when IP-Adapter scale is 0
+                print("[IP-ADAPTER] No faces - using dummy embeddings with scale=0")
+                
+                # Create zero embeddings of the expected shape (512D for InsightFace)
+                import numpy as np
+                face_embeddings = np.zeros(512, dtype=np.float32)
+                
+                # Disable IP-Adapter
                 try:
                     self.pipe.set_ip_adapter_scale(0.0)
-                    print("[IP-ADAPTER] Disabled (scale set to 0) - no faces detected")
+                    print("[IP-ADAPTER] Disabled (scale set to 0)")
                 except Exception as e:
                     print(f"[IP-ADAPTER] Could not disable: {e}")
             
@@ -485,8 +493,6 @@ class RetroArtConverter:
                 ]
                 
                 # Control guidance timing (when each ControlNet is active)
-                # [start, start] - both active from beginning
-                # [end, end] - both active until end
                 pipe_kwargs["control_guidance_start"] = [0.0, 0.0]
                 pipe_kwargs["control_guidance_end"] = [1.0, 1.0]
                 
@@ -503,6 +509,10 @@ class RetroArtConverter:
                 # Have keypoints but no embeddings (shouldn't happen but handle it)
                 print("Using keypoints only + Depth ControlNets (no embeddings)")
                 
+                # Create dummy embeddings
+                import numpy as np
+                face_embeddings = np.zeros(512, dtype=np.float32)
+                
                 # Control images: [face keypoints, depth map]
                 pipe_kwargs["control_image"] = [face_kps_image, depth_image]
                 
@@ -516,8 +526,10 @@ class RetroArtConverter:
                 pipe_kwargs["control_guidance_start"] = [0.0, 0.0]
                 pipe_kwargs["control_guidance_end"] = [1.0, 1.0]
                 
-                # DO NOT set ip_adapter_scale or image_embeds
-                print("  [WARNING] No face embeddings - using keypoints only")
+                # Pass dummy embeddings with scale=0
+                pipe_kwargs["image_embeds"] = face_embeddings
+                
+                print("  [WARNING] Using dummy embeddings - keypoints only")
             
             else:
                 print("No faces detected - using Depth ControlNet only")
@@ -530,15 +542,38 @@ class RetroArtConverter:
                 pipe_kwargs["control_guidance_start"] = [0.0, 0.0]
                 pipe_kwargs["control_guidance_end"] = [1.0, 1.0]
                 
-                # IMPORTANT: Don't pass ANY IP-Adapter related parameters
-                # No image_embeds, no ip_adapter_scale
+                # IMPORTANT: The InstantID pipeline REQUIRES image_embeds
+                # Pass dummy embeddings when no faces are detected
+                # Create zero embeddings of the expected shape
+                import numpy as np
+                dummy_embeddings = np.zeros(512, dtype=np.float32)
+                
+                # Ensure it's a numpy array with the right shape
+                if len(dummy_embeddings.shape) == 1 and dummy_embeddings.shape[0] == 512:
+                    pipe_kwargs["image_embeds"] = dummy_embeddings
+                    print(f"  [INFO] Using dummy embeddings (shape: {dummy_embeddings.shape})")
+                else:
+                    # Fallback to correct shape if something went wrong
+                    pipe_kwargs["image_embeds"] = np.zeros(512, dtype=np.float32)
+                    print("  [INFO] Using dummy embeddings (IP-Adapter disabled)")
+                
+                # Make sure IP-Adapter is disabled
+                try:
+                    self.pipe.set_ip_adapter_scale(0.0)
+                except:
+                    pass
             
             # Generate
             print(f"Generating: Steps={num_inference_steps}, CFG={guidance_scale}, Strength={strength}")
             
             # Debug: Print what IP-Adapter params we're passing
             if "image_embeds" in pipe_kwargs:
-                print(f"[DEBUG] Passing image_embeds: shape={pipe_kwargs['image_embeds'].shape}")
+                embeds = pipe_kwargs['image_embeds']
+                if isinstance(embeds, np.ndarray):
+                    is_dummy = np.allclose(embeds, 0)
+                    print(f"[DEBUG] Passing image_embeds: shape={embeds.shape}, dummy={is_dummy}")
+                else:
+                    print(f"[DEBUG] Passing image_embeds: type={type(embeds)}")
             else:
                 print("[DEBUG] No image_embeds in pipeline kwargs")