@Vision Pro Engineer Thank you for the reply. I think I have this figured out now. To answer your questions. The characters have both the Model and Skeleton components in RCP3. The entity itself is just a root node with a reference to the prototype (SkeletonMage). I'm exporting via the file menu File > Export Selected > Reality I believe the problem that I'm running into is that my findModelEntity method is exiting out too soon for reality files. The Model and Skeleton are expected to be on the second Rig_Medium. It does this correctly for the entity loaded via USD. For the entity loaded via Reality File, it sees that SkeletonMage is a ModelEntity and returns early. // output for usd file Rig_Medium is a model entity joint: head exists tree for SkeletonMage USD + + root + Rig_Medium + Rig_Medium + Skeleton_Mage_ArmRight + Skeleton_Mage_ArmRight + Skeleton_Mage_LegRight + Skeleton_Mage_LegRight + Skeleton_Mage_Eyes + Skeleton_Mage_Eyes + Skeleton_Mage_Hat + Skeleton_Mage_Hat + Skeleton_Mage_ArmLeft + Skeleton_Mage_ArmLeft + Skeleton_Mage_LegLeft + Skeleton_Mage_LegLeft + Skeleton_Mage_Body + Skeleton_Mage_Body + Skeleton_Mage_Skull + Skeleton_Mage_Skull + Skeleton_Mage_Jaw + Skeleton_Mage_Jaw // output for reality file Skeleton_Mage is a model entity target: Skeleton_Mage is a model entity but our joint is in another castle tree for SkeletonMage Reality + SkeletonMage + Skeleton_Mage + root + Rig_Medium + Rig_Medium + Skeleton_Mage_Hat + Skeleton_Mage_Hat I changed how I was searching for the rig and verified that the desired joint is present before creating the pin. And thanks for the tip on AttachedTransformComponent. You're right that I should be using that instead of a custom system. I had other plans for the PinnedEntitySystem that I just don't need at the moment.

Thank you. This was very helpful. I was able to find and fix my problems thanks to your comment. The issues preventing the skeletons from matching were that the rest transforms were different the joints were ordered differently between skeletons Notes One of the test characters had a (valid) subset of the expected bones. Fixed by adding the missing bones. Blender appears to be setting the rest position for the skeleton to the starting frame of the animation. This causes every exported animation to have a different restTransform. Fixed by correcting the rest transform after export. Blender does not export joints in alphabetical order. The order depends on an internal representation. Which means that even if the hierarchies match in Blender's outliner, they may not match in the exported USD. Fixed by reordering joints and related data after export so they all match. The rest pose and animations had the following joint order. I'm hoping the Blender issues are user error on my part, but I could not find any way to change the joint export order.

OK. Thank you for the reply @DTS Engineer . I'll namespace the conflicting model for now. Hopefully it won't be too difficult to migrate to the proper fix, if one is provided.

@joadan yes, the application crashes as well. The below view will terminated the app if you attempt to save a local item. struct ContentView: View { @State private var localId: Int = 0 @State private var localCount: Int = 0 @State private var remoteId: Int = 0 @State private var remoteCount: Int = 0 private var localContainer: ModelContainer private var remoteContainer: ModelContainer private let log = Logger() init() { localContainer = { do { let schema = Schema(versionedSchema: LocalSchema.self) let config = ModelConfiguration("local", schema: schema, isStoredInMemoryOnly: true, allowsSave: true, cloudKitDatabase: .none) return try ModelContainer(for: schema, configurations: config) } catch { fatalError("could not create local store") } }() remoteContainer = { do { let schema = Schema(versionedSchema: RemoteSchema.self) let config = ModelConfiguration("remote", schema: schema, isStoredInMemoryOnly: true, allowsSave: true, cloudKitDatabase: .none) return try ModelContainer(for: schema, configurations: config) } catch { fatalError("could not create remote store") } }() } var body: some View { HStack { VStack { Text("\(localCount)") Button("Save Local") { do { log.info("save local") localId += 1 let context = ModelContext(localContainer) let item = LocalSchema.Item(title: "local: \(localId)", created: .now, modified: .now) context.insert(item) try context.save() localCount = try context.fetchCount(FetchDescriptor<LocalSchema.Item>()) } catch { log.error("\(error.localizedDescription)") } } } VStack { Text("\(remoteCount)") Button("Save Remote") { do { log.info("save remote") remoteId += 1 let context = ModelContext(remoteContainer) let item = RemoteSchema.Item(title: "remote: \(remoteId)", created: .now, modified: .now, origin: "from space") context.insert(item) try context.save() remoteCount = try context.fetchCount(FetchDescriptor<RemoteSchema.Item>()) } catch { log.error("\(error.localizedDescription)") } } } } } }

I have created FB20099118 and the attached project still reproduces the issue.

Do the entities that you want to drag have colliders and InputTargetComponents? If so you can use the entity value on the tap and drag gesture directly. TargetedToAnyEntity requires an input target component to trigger. The below lets you tap on a box to change the material and drag a box to a different location. I believe both Tap and Drag gestures have a hitTest method that you can use as well. import SwiftUI import RealityKit struct ContentView: View { @State var root: Entity = Entity() @GestureState private var isDragging: Bool = false var body: some View { ZStack(alignment: .bottom) { RealityView { content in content.add(root) for idx in 0...3 { let ball = ModelEntity(mesh: .generateBox(size: 0.1), materials: [SimpleMaterial(color: .blue, isMetallic: false)]) ball.generateCollisionShapes(recursive: true) ball.components.set(InputTargetComponent()) ball.position.y = 0.3 * Float(idx) root.addChild(ball) } } Button("Reset") { for entity in root.children { guard entity.components.has(InputTargetComponent.self) else { continue } entity .components[ModelComponent.self]?.materials = [ SimpleMaterial(color: .blue, isMetallic: false) ] } } } .gesture(TapGesture().targetedToAnyEntity().onEnded{ tap in tap.entity .components[ModelComponent.self]?.materials = [ SimpleMaterial(color: .green, isMetallic: false) ] }) .gesture(DragGesture(coordinateSpace: .global) .targetedToAnyEntity().updating($isDragging) { value, state, _ in if let location3d = value.unproject(value.location, from: .global, to: .scene) { value.entity.setPosition(location3d, relativeTo: root) } }) } } #Preview { ContentView() }

I have created FB19426676. It includes an example project that reproduces the issue on my device.

You can use simple planes if you want but you have to account for the fact that the anchor is not necessarily the center of the plane. You will still get a "bounding box" style representation and there may be some over/under lap but it should be much closer to what you were expecting. Something like the following should work for you on both vertical and horizontal planes. The relevant line is the setTransformationMatrix entity.setTransformMatrix( anchor.originFromAnchorTransform * anchor.geometry.extent.anchorFromExtentTransform, relativeTo: nil) if generateSimplifiedPlanes { let mesh: MeshResource = try! await MeshResource.generatePlane( width: anchor.geometry.extent.width, height: anchor.geometry.extent.height) modelEntity.model?.mesh = mesh modelEntity.collision?.shapes = [try! await ShapeResource.generateStaticMesh(from: mesh)] entity.setTransformMatrix( anchor.originFromAnchorTransform * anchor.geometry.extent.anchorFromExtentTransform, relativeTo: nil) } else { let shape: ShapeResource = try! await ShapeResource.generateStaticMesh( positions: anchor.geometry.meshVertices.asSIMD3(ofType: Float.self), faceIndices: anchor.geometry.meshFaces.asUInt16Array()) modelEntity.collision?.shapes = [shape] modelEntity.model?.mesh = planeAnchorToMeshResource(anchor) entity.transform = Transform(matrix: anchor.originFromAnchorTransform) } Notes: It appears that the extent is always defined on a vertical plane so if you apply anchorFromExtentTransform you always want to use a vertical plane. This is rough code from a while ago and I haven't checked to see if there are newer/easier ways to do this.