Cloud Anchors, Geospatial AR, and a Full Plane Detection Rewrite: What’s New in ViroReact 2.53.0
Today, we are launching three major updates simultaneously: ViroReact 2.53.0, Editor Alpha 3, and Platform Public Beta. Together, these three releases represent a huge milestone for ReactVision and a major step towards our goal of building a complete end-to-end stack for spatial apps.
You can read about the Editor and Platform updates in their own announcement posts. In this post, we are going to focus on what is new in ViroReact 2.53.0 and, more importantly, why it matters.
ReactVision is Now Your Default AR Backend
This is the one I am most excited about. Since introducing Cloud and Geospatial anchors, ViroReact has relied on Google’s ARCore API for its persistent and location-based AR features. With 2.53.0, that changes. ReactVision’s platform is now the default provider for both Cloud Anchors and Geospatial Anchors. The provider prop on ViroARSceneNavigator defaults to "reactvision" .
No Google Cloud account required. No ARCore Geospatial API setup. No VPS. Just your ReactVision Studio API key and project ID, and you are up and running.
The existing hostCloudAnchor, resolveCloudAnchor, and geospatial anchor APIs you already know are unchanged, so the migration path is simple. And if you still need Google's ARCore provider for a specific use case, you can explicitly set provider="arcore" and everything works exactly as before.
Cloud Anchors Get a Full Management Layer
One of the biggest limitations of the previous cloud anchor integration was that it was essentially fire-and-forget. You could host and resolve anchors, but managing them after the fact was largely invisible.
With ReactVision Cloud Anchors, we have added eight new methods on arSceneNavigator that give you full control: list all your project's anchors, fetch individual records, update names and descriptions, delete anchors you no longer need, and even search for anchors near a GPS location. You can also attach assets directly to anchors (more on that below) and track resolution analytics.
This turns cloud anchors from a basic feature into a first-class, manageable part of your application’s infrastructure.
Geospatial Anchors Without the Google Dependency
Similarly, geospatial anchors are now fully supported through the ReactVision provider. You can create, host, resolve, list, search, update, and delete geospatial anchors entirely through our platform. createGeospatialAnchor, createTerrainAnchor, and createRooftopAnchor all work with provider="reactvision".
We have also added two new utility functions, gpsToArWorld and latLngToMercator, that make it easier to convert GPS coordinates into AR world-space positions. These are handy building blocks whether you are placing content at a specific location or building custom geo-spatial logic.
Asset Uploads: Connecting Content to Anchors
A new rvUploadAsset method lets you upload 3D models, images, video, and audio directly to ReactVision storage from within your app. The returned asset ID can then be linked to either a cloud anchor or a geospatial anchor.
This is especially powerful for user-generated content scenarios. Imagine an app where users can create a 3D model (perhaps using AI, like in Play AR), upload it, and pin it to a real-world location for others to discover. The upload, storage, and anchor association all happen through a single, consistent API surface.
Shader Modifiers: Even More Power
In 2.52.0, we introduced custom shaders. In 2.53.0, we are expanding them significantly.
You can now pass custom textures to your shaders and swap them at runtime. Vertex-to-fragment data passing via varyings is supported, so your geometry modifiers can drive your surface modifiers. Fragment modifiers can access the scene depth buffer for effects like soft particles and contact glow. And you can sample the live camera feed directly onto any surface for portal, refraction, and magnifying-glass effects.
We have also added a priority field for deterministic ordering, so your custom effects always layer correctly regardless of attachment order.
AR Plane Detection: Rebuilt from the Ground Up
ViroARPlaneSelector has been completely rewritten. The old implementation had a number of long-standing issues: ghost planes, objects snapping to the wrong position, planes not cleaning up properly, and updates being silently dropped for large surfaces.
All of that is gone. The new implementation uses a scene-event-driven architecture and fixes every one of those bugs. Objects now land exactly where you tap, not at the plane centre. Plane detection defaults to both horizontal and vertical. And the adaptive update rate means you get fast initial detection without overwhelming the system.
This is a breaking change (you now forward anchor events via a ref), but the result is dramatically more reliable.
Stability and Bug Fixes
Beyond the headline features, we have continued our focus on stability. Overexposed 3D models, broken iOS video recording on the New Architecture, Android physics crashes, and several other issues have all been resolved. Check the full release notes for the complete list.
The Bigger Picture
If you step back and look at what 2.53.0 represents alongside the Editor and Platform updates, the vision is becoming clear. ReactVision is building the full stack for spatial applications: a powerful open-source renderer, a cloud platform for persistence and storage, and a visual editor for rapid prototyping.
This release is the moment those pieces start to connect. Your AR scenes can now talk to our cloud, store assets, manage anchors, and do it all without leaving our ecosystem you already know.
We are just getting started, and we cannot wait to see what you build.
Need Help?
Looking for support getting started with ViroReact or want an expert to help you build a comprehensive AR or VR application? Look no further than our community Discord and our trusted expert partners.