Is the video uploaded to a server?

No. Decoding, depth estimation, the stereo warp and the encode all happen on your device. Nothing is sent anywhere, which also means no queue and no per-minute limit.

Why is converting a video slow?

Every frame needs its own depth pass, so cost scales with frame count, not file size. On a GPU-backed browser the depth pass costs roughly 0.2 seconds per frame at 518x518; setting the depth refresh to every second frame halves that again.

What format does the output use?

Half-width side-by-side (SBS) H.264 in an mp4 container, with the original audio re-encoded to AAC. That is the layout every major headset player accepts.

Why does my SBS video look stretched or doubled?

Usually the player is not in 3D SBS mode, or the display is showing the raw frame. Switch the player to SBS/3D and use half-SBS output; full-SBS doubles the width and many players then letterbox or crop.

← DepthSnap / Guides / Video to 3D SBS

Video to 3D SBS: convert any clip in your browser

Q: Can I convert a video to 3D SBS for free?

Yes. DepthSnap converts clips in the browser with no upload, no signup and no watermark. It runs an AI depth model on every frame locally and encodes a side-by-side H.264 mp4 you can copy to a headset.

Photos take one depth pass. Video takes one per frame — that single fact decides everything about speed, quality and which tool you should use.

What actually has to happen

Decode a frame at the output resolution (browser video element or WebCodecs).
Estimate depth for that frame with a monocular model — we use Depth Anything ViT-S at 518×518.
Warp the frame into two views using the depth map: near objects get crossed parallax, far objects diverge behind the screen plane.
Encode the side-by-side composite as H.264 and mux the original audio back in as AAC.

Steps 1, 3 and 4 are cheap (single-digit milliseconds per frame). Step 2 is 95% of the wall clock, and it is why "1-minute clip" can mean anything from 30 seconds to an hour depending on the machine and the browser.

How to do it

Open the video to 3D converter and drop in an mp4, mov or webm clip — or hit Try a sample clip if you just want to see the result first.
Pick the output frame rate (12 or 24 fps) and the depth refresh rate. Every 2nd frame is the sweet spot for most footage: half the work, no visible difference on slow camera moves.
Press convert, wait for the progress bar, download the -sbs.mp4 and copy it to the headset.

Short clips convert dramatically better: 10–30 seconds is the sweet spot for social video, demo reels and highlights. If you need a long film, cut the scene you care about first.

Real timings (measured, not marketing)

Numbers from a browser on Apple silicon, half-SBS output, 640×360 per eye, depth refresh every 2nd frame:

Setup	Depth pass / frame	30 s clip @24 fps
WebGPU (Chrome/Edge, recent GPU)	≈ 0.18–0.25 s	≈ 2–4 minutes
WASM, several threads	≈ 1.0 s	≈ 10–14 minutes
WASM, single thread (older browsers)	≈ 3–5 s	an hour or more

Two conclusions follow. First, use a browser with WebGPU — Chrome, Edge, or the Chromium browser on Quest. Second, prefer short clips and the "every 2nd frame" depth setting; together they cut the time by about 4× versus the worst case.

Playback on each headset

Device	What to use	Notes
Meta Quest 2/3	Files app / Skybox / DeoVR	Choose "3D side-by-side". If it looks doubled, the player is still in 2D mode.
Apple Vision Pro	Files → spatial video, or Moon Player	SBS mp4 plays as 3D when the player is set to side-by-side.
Pico 4	Pico Video	Switch layout to SBS; half-SBS keeps the aspect ratio correct.
XREAL / VITURE glasses	Any player on the phone or PC driving them	Set the display to SBS 3D; glasses act as a dumb display.
3D TV / projector	USB playback	Pick "side-by-side" in the TV's 3D menu.

Half-SBS (each eye half the frame width) is the layout with the widest support. Full-SBS doubles the width and is often downscaled or cropped by players — see Half vs Full SBS.

Why the result can look wrong

Stretched or doubled image — the player is not in SBS mode, or you exported full-SBS and the player assumed half. Fixes in this guide.
Depth that feels flat — disparity is being clamped. A good converter scales parallax from the depth range (a few percent of the eye width), not a fixed pixel shift.
Wobbling edges — depth changing every frame. Lower the depth refresh rate for slow footage; raise it for fast camera moves.
Audio out of sync — the muxer must be told the real sample rate of the decoded track, not a guessed 48 kHz.

Browser converter vs desktop tools

	Browser (DepthSnap)	Desktop converter
Install	None	Installer, often licensed
Upload / privacy	Nothing leaves the device	Usually none either, but installs a binary
Cost	Free	Often paid / watermark
Speed on long clips	Minutes for short clips	Faster on long renders
GPU requirement	WebGPU strongly recommended	Any CUDA/Metal GPU

FAQ

Can I convert a video to 3D SBS for free?
Yes — the video converter is free, has no watermark and needs no account.

Is my video uploaded?
No. Everything runs locally in the browser tab.

Why is it slow?
One AI depth pass per frame. Drop the depth refresh to every 2nd frame, or use a shorter clip.

What exactly do I get back?
A half-SBS H.264 mp4 with AAC audio, playable on any headset 3D player.

Does it work on Safari?
It runs, but without WebGPU the depth pass falls back to WebAssembly and is several times slower. Chrome/Edge on the same machine is much faster.

Convert a video to 3D SBS →

DepthSnap — free browser-based 2D to 3D SBS converter for photos and video. No upload, no signup, no watermark. Independent tool; device trademarks belong to their owners.

More guides: Half vs Full SBS · Stretched SBS fix · SBS vs TB · per-device setup