360 Gaussian Pro Turns 360 Video Into 3D Splats Entirely on Your PC

Developer Antti Ahokoivu shared the new Pro workflow on August 17. The demonstration moves through a reconstructed alley as if it were a small virtual location, then shows the controls used to process the source footage.

Four steps pull the tools together

The app begins by pulling frames from raw dual-fisheye footage, a stitched 360 video, or even an ordinary flat video. It can keep the sharpest frame from each interval, which gives the reconstruction cleaner material without saving every moment of the clip.

Masking comes next. Meta’s SAM 3 model can find something from a word such as person, while fisheye presets remove the empty area outside each circular image. Creators can also paint a mask by hand when the automatic result needs help.

The alignment step uses a custom build of COLMAP with GLOMAP support. In plain language, it compares the overlapping images, works out where the camera was for each one, and builds a sparse map of the scene. Extra photos from a phone, DSLR, or drone can be added when one part of the location needs more detail.

Finally, 360 Gaussian Pro can pass that aligned data to LichtFeld Studio, Postshot, or Brush for training. It can also leave the images, masks, camera positions, and sparse point cloud ready for another Gaussian splat trainer.

A Gaussian splat is different from a traditional polygon model. It represents a captured place with many soft, colored points in 3D. Those points can be drawn quickly enough for someone to move around the scene in real time, which is why the format has become so useful for virtual production and spatial experiences.

The footage stays on your computer

Every processing step runs locally, and 360 Gaussian Pro says nothing is uploaded. That is especially useful when a scan contains an unreleased set, a private interior, or simply a lot of high-resolution camera footage.

Ahokoivu built the software for fast, batch-processed scans of filming locations. That origin makes sense when you see the queue and the adjustable settings. A cinematographer can capture a real place, process several clips overnight, and later explore camera ideas without returning to the location.

The app supports raw footage from Insta360 X4, X5, and X6 cameras, DJI Osmo 360, Antigravity A1, and DJI Avata 360. It also accepts stitched equirectangular video, regular video, and still photos from other cameras.

It requires 64-bit Windows 10 or 11 and an NVIDIA RTX 20-series GPU or newer. The developer recommends 12 GB of video memory, at least 16 GB of system memory, and NVIDIA driver R580 or later. AMD and Intel graphics are not supported.

You can try the whole pipeline for free

The free trial includes every step and does not require an account, but it is limited to 14 frames. A one-week license costs €20 without automatic renewal, one year costs €89, and a one-time v1.x license costs €149. Those prices include VAT.

The frame limit makes the trial more of a proof than a full location scan, but it is enough to see whether the workflow runs on your computer and how the pieces fit together.

360 Gaussian Pro is available to download from the developer’s website (opens in a new tab). The best part is how ordinary the starting point can be. A clip from a small 360 camera can become a place you move through, scout, or carry into another 3D tool. The app makes that jump feel much closer to pressing Run Pipeline than building one from scratch.