3D Lab

Video to 3D

Turn an object video into a 3D scan using multiple views. Join the private GPU queue and download a colored GLB model. Experimental.

Shared processing queue

Checking availability…

One photo Object video

Film close enough to see the detail

Keep the whole object visible and let it fill roughly two-thirds of the picture height. Move the camera slowly around a stationary object, with overlapping views of every side. Use even lighting, a normal lens and a textured, matte subject. Avoid changing zoom, blur, reflections and moving hands.

Small objects far from the camera lose detail when video frames are sampled. A full turn with clear views of the front, back and sides works better than repeating one angle.

Experimental multi-view scan · 5–90 seconds · up to 200 MB locally. Your browser sends 36 sampled images, up to 960 pixels per side; the original video and audio stay on your device. Reconstruction can take up to 30 minutes after it starts.

Choose an object videoMP4, WebM or a browser-supported MOV · 5–90 seconds

The scan uses overlapping views and captured colors. When surface recovery is needed, outlines help estimate missing shape; small details and concavities can be wrong. Unusable scans may fail quality checks. Scale is relative; inspect every side before using the model.

Email delivery is not available yet. Return to this page in the same browser to check your saved job; keep site cookies enabled.

I have the right to use this video and agree to send sampled images to MakePip’s private processing worker. Files are temporary. Download the result within 24 hours after completion.

Queue wait: up to 12 hours to start. If it cannot start in that time, the request expires. Results remain available for 24 hours after completion.

How to use Video to 3D

1Film a slow walk around one stationary, textured object, with steady light and overlapping views.
2Choose a 5–90 second clip. Your browser samples 36 frames; confirm processing and join the queue.
3Inspect the reconstructed surface and captured colors, then download the GLB within 24 hours.

Before you start

Does it use the whole video?

The browser samples 36 views throughout the clip. The processor matches features between those images, estimates camera positions and reconstructs depth and a surface. It does not just turn one selected frame into an AI model.

Is my original video uploaded?

No. Only sampled images are sent after confirmation. The original video and its audio stay on your device. Sampled frames are limited to 960 pixels per side and a 12 MB upload.

Which recordings work best?

Keep one matte, textured object still and walk around it slowly using a normal lens. Maintain overlap and even lighting. Avoid moving hands, changing zoom, motion blur and reflective, transparent or plain white subjects.

How long does it take?

Queued requests wait for the private processor to have capacity. A scan can take several minutes and has a 30-minute processing limit. Unstarted requests expire after 12 hours.

Is it a dimensionally accurate scan?

The result has relative scale and can contain holes or missing surfaces. It is not calibrated for measurements or verified for 3D printing.

Read the guides →File processing & privacy →