SamuSynth Logo Vid2Scene Beta

Upload a Video to Generate a 3D Scene

Because video files are usually large, we suggest connecting to WiFi before uploading.
Want to make sure you get great results?
See an example video recording or check the tips below.

  • Capture from Multiple Angles: Try to capture all subjects in the scene from multiple angles. Otherwise, the subjects may look incorrect from certain angles or viewpoints.
  • Use Lateral Movement: Try to move the camera laterally (left, right, up, down) to capture the scene from different angles. Having only forward or backward movement will cause the scene to look incorrect from certain angles or cause a failed reconstruction.
  • Capture only one area: Try to capture only one area per reconstruction (i.e. one room, one outdoor space, etc.). Too many areas in one reconstruction may cause it to fail.
  • Ensure Good Lighting: Make sure the area is well-lit to capture clear visuals.
  • Keep your Movements Smooth: Avoid moving the camera too quickly to prevent motion blur. Also, try not to rotate the camera too quickly either -- it may cause certain frames to be missed during the reconstruction process.
  • What's the ideal length? The ideal video length is between 1 to 3 minutes.

Max file size: 450MB

Check this if your video was captured with a 360° camera and is in equirectangular (360°) format.

Note: Scenes uploaded without logging in will be publicly accessible using a direct link.
Login to keep track of all of your scenes or to create private scenes.

We'd love to showcase your scene as an inspiring example for other users! By checking this box, you're giving us permission to feature the scene on our examples page if selected.

Apply a filter to all frames before processing. This can give your scene a unique look.

Min: 100,000 Max: 1,200,000

Controls the maximum number of points in your 3D scene. Higher values create more detailed scenes but require more processing time.

Min: 5,000 Max: 30,000

Controls how long the training process runs. More steps generally improve quality (with diminishing returns) but increase processing time.

This removes the background from the video before processing. This is useful if you want to scan a specific object instead of a whole area.

Generates a sphere of points around your scene that's used during the gaussian splatting training process. It can help remove floaters and make the background look more detailed, especially for outdoor scenes.

More coming soon!

Estimated processing time: 1.5 to 3 hours