Skip to content
360 Gaussian Pro logo360 Gaussian Pro

Input formats

What kinds of video and photos 360 Gaussian Pro accepts, and how it treats each.

360 Gaussian Pro accepts several input types and picks the right pipeline for each automatically when you add a clip to the queue. You can also adjust settings manually.

Supported inputs

Input Examples Notes
Dual-fisheye 360 Insta360, Antigravity A1 .insv, DJI Osmo 360, Avata 360 .osv The highest-quality path especially for interior splats. Both lenses are used natively.
Equirectangular (stitched 360) 2:1 .mp4 / .mov Single stitched sphere.
Flat (non-360) video Smartphone / DSLR / drone .mp4 / .mov Ordinary perspective video — the frames are used directly.
GoPro video (experimental) HERO5 and newer, MAX (single-lens) .mp4 Treated as flat video, plus the camera's gravity and GPS data.
Photo folders DSLR / drone stills A folder of plain photos.
Input formatsPick an input to see the path it takes
Extract→Pinhole split→Mask→Align→Train

Both lenses used natively — the quality path, especially for interiors.

.insv · .osv · GPS / IMU used

How input type changes things

  • Dual-fisheye footage is split into pinhole views and fused into a single camera rig — the quality path for 360 capture.
  • Equirectangular footage is aligned as 360 footage and split later in the process into cubemaps.
  • GPS / IMU if footage has GPS or IMU data embedded in .insv / .osv files is used automatically to put the model upright and at real-world scale. If footage is trimmed outside of 360 Gaussian Pro, it's not able to extract GPS data. You can provide a .gpx file next to the input file. For inputs without embedded GPS/IMU, the model is reconstructed in an arbitrary orientation and scale — you can align and rescale it manually afterwards.
  • GoPro .mp4 clips carry gravity and GPS in their metadata. The app reads them, so the model comes out upright and at real-world scale. GPS is used only when the camera had a satellite lock for almost the whole clip. Without embedded GPS, a .gpx file next to the video is used. GoPro MAX .360 files are not supported, and rolling shutter correction isn't applied to GoPro footage.
Which input suits which scenePick what you are shooting
InputDual-fisheye 360
Splits5
WhyTight rooms need every direction at once; the native fisheye path keeps the most detail.
Watch forMask the operator — you are always in shot.

Batches

A batch is one project built from several files. Use it to combine several 360 takes of one space, and to add detail material: 360 stills, drone or phone videos, and DSLR or drone photos.

  1. Click Batch under the queue. An empty batch box appears with the text Drop videos, photos or folders here.
  2. Drag files or whole folders onto the box, or click + on the box and pick Add files… or Add folder…. Folders are searched with their subfolders. You can keep adding to the box later.
  3. The app sorts every file into its role by itself:
You add… Role in the batch
360 videos (.insv, .osv, or stitched 2:1 .mp4 / .mov) 360 videos. They build the model and are aligned as one scene.
Raw .insp photos (with Insta360 .insv videos) 360 stills, from the same camera
Stitched 2:1 .jpg / .png panoramas (with stitched 360 videos) 360 stills
Ordinary videos (drone, phone, DSLR) Detail videos
Ordinary .jpg / .png photos Detail photos, in sets

The box name shows what it holds, for example Batch (2 videos + 1 detail video + 1 photo set). Click × on a member row to take it out of the batch.

What the 360 videos do

The 360 videos alone build the model. Detail material never changes the main alignment. It is added to the finished model after Align, and only adds sharpness and coverage. If some detail images fail to register, the run still completes.

  • One batch holds one kind of 360 video: all dual-fisheye or all stitched equirectangular. Adding the other kind is refused.
  • A box without a 360 video shows Batch — needs a 360 video and can't run yet. Drop a 360 video into it.
  • A batch of only ordinary videos is a normal flat-video project (no 360). It can't take detail photos.

360 stills

Raw .insp photos carry much more resolution than the video frames (11904×5952 on the X5), so they add sharp detail where you stopped to take a photo. Take them with the same camera as the videos, because they use its lens calibration. DJI .osv batches take no stills.

The stills appear in the Select batch clip list on the Split tab as Stills (N). Use the trim handles there to choose which ones are used.

Detail videos and photos

  • Detail videos appear in the Select batch clip list with (detail) after the name. Each has its own trim and frame rate. Trim it to the part that overlaps the 360 footage, such as a drone's low climb-out.
  • Detail photos come in sets, one set per camera. A folder you add is one set. Loose photos are grouped by camera model, focal length and size from their EXIF data. Don't zoom or mix cameras inside one folder. Sets show as folder/ (N photos) in the same list. All photos in a set are used.
  • Mask detail frames and stills on the Masking tab like any other frames. Pick the member in the list. Detail members never get the 360 lens mask.

For good registration, shoot the detail material during the same visit and from positions near the 360 camera path. Step gradually from the 360 path toward the detail subject. Images with no visual overlap with the 360 footage won't register. Photos must be .jpg or .png; convert HEIC and RAW to JPG first.

Adding detail to a finished project

You can also drop detail videos or photos onto a single 360 clip in the queue. It stays the same project, now with detail members. Dropping a second 360 video turns it into a new batch project.

If the project is already extracted, only the new detail material is extracted. If it is already aligned, the detail is registered right away, with no need to run Align again.

Dragging files onto empty space in the queue never creates a batch. Each video becomes its own clip, and a photo folder becomes a photo project. To combine them, drop them onto a batch box.

Next: First Run.