/

OpenQuestCapture

Social Media Links

Github

A Hyperscape alternative that lets you capture with your Quest 3 or 3S.

Location

Size

Total raised

Hiring?

Platform

Quest 3 and Quest 3S app, sideloaded from SideQuest or from the APK on the GitHub release. The companion reconstruction scripts run on desktop Python.

GPU

None for capture, which runs entirely on the Quest 3 or 3S. The companion quest-3d-reconstruction pipeline builds a COLMAP model on CPU, and training the resulting scene needs whatever GPU your Gaussian splatting trainer requires.

Pricing

Free

License

MIT

Best for

Capturing Quest 3 passthrough stereo, depth and pose data for offline 3DGS reconstruction

Updated

July 2026

OpenQuestCapture is a Unity based data logging and visualisation app for the Meta Quest 3 and Quest 3S, built on top of the Quest sensor access layer from t-34400's QuestRealityCapture project. It records synchronised headset and controller poses, raw YUV passthrough frames from both cameras, Camera2 API intrinsics and characteristics, and per eye depth maps, writing each run into a timestamped folder on the headset's internal storage. The first prebuilt APK appeared in December 2025 and the current build is v1.0.2, which fixed a fault where the app silently stopped capturing images after the headset woke from sleep. The repository carries 195 stars and 37 forks.

Recording starts and stops from the Menu button on the left controller, and the Y button opens a session manager that lists past captures, exports them to a zip in the Quest's Download folder and deletes them to reclaim space. While recording, a depth point cloud overlay shows coverage in place, with each point tinted by the angle it was observed from, so white marks head on coverage and saturated tints mark grazing angles. Depth and camera frames are captured at three frames per second by default, with timestamp alignment held between the two streams through a shared monotonic and Unix time base.

Reconstruction happens off the headset. The companion quest-3d-reconstruction repository ships an end to end script that converts the YUV frames to RGB, rebuilds a coloured point cloud and writes a COLMAP sparse model of images, cameras and points that any Gaussian splatting trainer can consume. An optional tone mapping stage combining CLAHE and gamma correction is exposed in the pipeline configuration, aimed at interiors where window light blows out the highlights and crushes the shadows in the same frame.

The README still names a hosted alternative to that local pipeline, and that route no longer exists. vid2scene ceased website operations on 13 June 2026, so its Quest upload endpoint no longer accepts session zips and there is no managed cloud path for OpenQuestCapture data. The vid2scene codebase was released under Apache 2.0 by the same author, which leaves the same reconstruction available to anyone willing to run it themselves.

The app is MIT licensed and free, and it redistributes Meta's OpenXR SDK, so a derived build has to honour that licence separately. Headset camera access has to be granted at first launch or no images are recorded at all, which is the most common reason a session comes back with poses and depth but no imagery. The author describes the project as early and in active development, and names export speed as the known weak point, since writing a session to zip can take several minutes.

Frequently Asked Questions

Which headsets does OpenQuestCapture support?

OpenQuestCapture runs on the Meta Quest 3 and Quest 3S only. It is built against Unity 6000.2.9f1 with the Meta OpenXR SDK and the Mixed Reality Utility Kit, and it installs by sideloading, either from SideQuest or from the APK attached to the GitHub release.

Does OpenQuestCapture produce a Gaussian splat by itself?

OpenQuestCapture does not train splats. It records headset and controller poses, raw stereo passthrough frames and per eye depth maps, and the companion quest-3d-reconstruction scripts turn that data into a COLMAP sparse model that a separate Gaussian splatting trainer consumes.

Can OpenQuestCapture data still be uploaded to vid2scene?

No. vid2scene ceased website operations on 13 June 2026 and its Quest upload page is gone, although the OpenQuestCapture README still lists it as the recommended route. The vid2scene codebase is Apache 2.0 on GitHub and can be run locally instead.

What licence does OpenQuestCapture use?

OpenQuestCapture is released under the MIT licence and is free to download. It also redistributes Meta's OpenXR SDK, so anyone shipping a derived build has to comply with Meta's terms separately.

How long should an OpenQuestCapture session be?

OpenQuestCapture sessions work best between one and three minutes on the author's own guidance. The app records depth and camera images at three frames per second by default and draws a live depth point cloud so coverage gaps are visible before the session ends.

Alternatives

Recent Headlines