2026-09-14

iPhone spatial video costs you 130MB a minute for an effect almost no one sees (and what NearVid can still pull from the file)

Apple's own newsroom page states the cost plainly: "a minute of spatial video is approximately 130MB." The same page states, just as plainly, what that megabyte-hungry minute actually looks like almost everywhere it's opened: "On iPhone and other devices, spatial videos appear as regular, 2D videos and can be watched or shared just like any other video using apps like Messages." Put those two sentences together and you get an odd situation — a video format that costs roughly double the storage of an ordinary clip, for an effect that only one device on the market actually shows you. Here's what's really in that file, why it's that size, and which of NearVid's actual on-device tools can do something useful with it once it's sitting on your phone.

What "spatial" means, and which iPhones actually shoot it

Spatial video capture launched with iOS 17.2 in December 2023, on the iPhone 15 Pro and 15 Pro Max only — Apple's own announcement ties it directly to viewing on Apple Vision Pro, "life-size and with incredible depth." It didn't stay Pro-only for long: by the iPhone 16 launch in 2024, UploadVR reported that "all iPhone 16 models can capture spatial photos and spatial videos," using what Apple calls the Apple HEVC Stereo Video Profile — calling it "now a standard iPhone feature," not a Pro exclusive anymore.

It also isn't improving much. When the iPhone 17 lineup shipped in September 2025, UploadVR found spatial video on the iPhone 17 Pro "still limited to 1080p 30fps recording... matching the capability introduced with the iPhone 15 Pro in 2023" — unchanged across three generations, even as ordinary 2D video on the same phone jumped to 4K at 120fps. The iPhone Air, with its single rear camera, can't record spatial video at all; you need two lenses working together to capture two eyes' worth of footage.

Why one minute is 130MB: it's a second camera feed, not extra pixels

That "two eyes" part is the actual reason for the file size, and it has nothing to do with resolution. Spatial video is captured at a fairly modest 1080p, 30fps, standard dynamic range — Apple's own words, from the same newsroom page quoted above. What makes the file large is the format underneath it: Mike Swanson's technical write-up on the format explains that Apple spatial video uses MV-HEVC (multiview HEVC) to store "both left- and right-eye views" — two genuinely separate, full video streams, encoded together into one file, using a special metadata box (Apple calls it "vexu") so a compatible player knows to treat them as a stereo pair. It's not one 1080p video rendered more expensively; it's two of them, recorded by two cameras at once, riding inside a single .mov.

The part Apple says out loud: almost nothing plays back the second eye

This is the detail that makes the format genuinely interesting rather than just "big file, why bother." MV-HEVC is deliberately built so that a player which has never heard of spatial video still works: per Swanson, a standard, spatial-unaware decoder reads only the first, base view — the "left-eye" stream — and silently ignores the metadata and the second view sitting next to it in the same file. It doesn't error out or refuse to open; it just plays a normal video. That's exactly what Apple's own quote above describes happening "on iPhone and other devices" — the fallback isn't a workaround someone bolted on, it's the format working as designed.

The practical result: unless the exact file lands on an Apple Vision Pro (or the small set of other spatial-aware software that's shown up since), what actually gets watched — in Photos, in Messages, in Files, in NearVid, in nearly every video app on earth — is the ordinary flat half of the recording. The second eye is along for the ride, silently doubling the file's weight, doing nothing for the video's only realistic audience.

What NearVid can actually do with one of these files

None of NearVid's tools know what "spatial" means, and that turns out to be fine, because none of them need to — every one of them either works at the container level (no decoding at all) or decodes the same way any other ordinary player does, which per Apple's own description above is already the entire realistic audience for this footage:

• Extract Frame. Pull one timestamp out as a JPG or PNG. Grabbing a still only requires demuxing and decoding a single frame — the same base view every non-Vision-Pro app already shows — so a moment from a giant two-eye recording comes out as an ordinary photo you can actually text someone, no headset required.
• Extract Audio. Spatial video records Spatial Audio alongside the two video streams; pulling just the sound out as an MP3 (libmp3lame) never touches either video layer at all.
• Trim. NearVid's trim is a lossless stream copy — it cuts the container down to the seconds worth keeping without decoding or re-encoding anything, which matters here more than usual, since spatial footage is burning through storage at roughly double the rate of an ordinary clip of the same length while most of it never gets used.
• Convert to WebM. Re-encoding decodes first and encodes second — the decode step reads the same base view as everything else, so the WebM you get out is the flat video the file was already showing everyone but Vision Pro, just now a small, ordinary, freely shareable file instead of a fixed ~130MB-a-minute container.

The one honest gap: we haven't tested this against a real spatial-video export

Everything above follows from how MV-HEVC is documented to behave — but NearVid hasn't specifically been run against an actual Apple spatial-video .mov file, and the format's multiview metadata box is newer than a lot of the software that handles video containers. In practice that means one of two things happens when you drop one in: either NearVid's FFmpeg-based tools open it and work with the base view exactly as described above, or the demuxer trips on a box structure it doesn't recognize and the upload gets rejected outright. We'd rather say that plainly than imply this has been verified when it hasn't — if a given spatial-video file doesn't load, that's this exact, known gap, not a mystery bug, and it's worth testing on a file you don't mind losing before you trust the trimmed copy as your only one.

What none of this is

Nothing here creates spatial video, previews it in 3D, or does anything at all with the second eye — NearVid has no stereo viewer and isn't trying to build one. It doesn't turn an ordinary flat clip into a spatial one either. Every operation above only ever touches the same single, ordinary, already-flat view that Photos, Messages, and every other non-Vision-Pro app on the planet is already showing you when you open the file — NearVid just gives you a way to cut it down, pull a photo or an MP3 out of it, or shrink it into a normal format, entirely on your own device, without that ~130MB-a-minute original ever leaving it.

The honest summary

A minute of Apple spatial video is about 130MB, per Apple's own newsroom page, because the format — MV-HEVC — records two full video streams instead of one, per Mike Swanson's technical write-up on it. It's been standard on every iPhone 16 and newer model with two rear cameras since 2024, per UploadVR, and still shoots at only 1080p/30fps on the iPhone 17 Pro three years after launch, per UploadVR's coverage of that phone. And by Apple's own admission, on every device except Vision Pro, all that extra data plays back as nothing more than an ordinary 2D video. NearVid can't touch the spatial part of the file and was never built to — but trimming it losslessly, pulling a frame or the audio out of it, or converting it down to WebM all work with exactly the ordinary flat video the file was already showing you, on your own device, without uploading a 130MB-a-minute file anywhere to find out.