360 video is the cheapest way to make something that genuinely works in a headset. No engine, no code, no real-time rendering — a camera, an editor, and an export. It’s also the format people get wrong most often, because almost every habit from normal filmmaking is either useless or actively harmful here.
1. What the camera actually does
A 360 camera is two or more wide lenses pointing in opposite directions. Their images are combined into a single frame in a process called stitching, usually now done in the camera, sometimes in software afterwards.
The output is one video in equirectangular projection — the whole sphere unwrapped into a rectangle, the way a world map unwraps a globe, with the same distortion at the poles. It looks wrong flat and correct in a viewer.
Stitch lines are where the lenses meet. Anything crossing that boundary can smear or double, and the seam is usually along the sides. Keep important action away from the stitch line, which in practice means pointing one lens at whatever matters.
2. You cannot frame. You can only direct attention.
This is the real shift. There is no frame, no off-screen, and no cut to a close-up. The viewer looks where they want.
What still works:
- Motion and sound pull the eye. These are your two reliable tools.
- Position is your composition. You can’t choose the framing, but you choose where the camera stands — and in 360 the camera’s position is the viewer’s position. Put it where a person would plausibly be, at roughly eye height for a standing or seated viewer, not on a tripod at waist height, which makes everyone feel like a child.
- Give people time. Cuts that work in flat film are disorienting here. Shots need to be long enough to look around — often several times longer than feels right in the edit.
- Reset attention before you cut. If a viewer is facing backwards when you change scene, they’ll miss the start of the next one.
3. Hide, and mind the tripod
You’re in the shot. Everyone is in the shot. Standard practice is a thin monopod (which the camera can largely erase from the nadir), and the crew hiding behind something, around a corner, or removed in post from one of the two lens views.
4. Audio is where the immersion lives
Flat stereo ruins 360 video faster than mediocre picture does. If a viewer turns their head and the sound doesn’t move, the whole thing collapses into a screen wrapped around their face.
Ambisonics is the fix: a format that records the whole sound field rather than fixed channels, so it can be rotated to match head movement at playback. Practically:
- Record with an ambisonic microphone (a tetrahedral array) placed as close to the camera as you can hide it — it must share the viewer’s position, or sound and image will disagree.
- You’ll be working in first-order B-format: four channels called W, Y, Z, X. W is overall pressure; the other three are directional. You don’t need the maths, but you do need to keep the channel order straight, because getting it wrong produces audio that rotates the wrong way and is maddening to diagnose.
- YouTube accepts first-order ambisonics as a 4-channel track, or ambisonics plus head-locked stereo as 6 channels (W, Y, Z, X, L, R). Head-locked is for anything that shouldn’t rotate — narration, score.
Mixing happens in a DAW with a spatial plugin suite; Reaper is the common choice because it handles arbitrary channel counts without complaint, and DaVinci Resolve and Premiere both support ambisonic tracks in the timeline.
5. Export, metadata, and the thing that always goes wrong
The failure everyone hits once: the file plays as a distorted rectangle instead of a sphere. That’s missing spherical metadata — a tag telling the player this is 360 footage. Modern editors have a checkbox for it; if yours doesn’t, there are injector tools that add it after export.
Export high. 360 spreads resolution across the whole sphere, so the viewer sees only a fraction of your pixels at any moment. 4K sounds generous and looks soft in a headset; 5.7K or 8K source is the norm for a reason.
6. A first project
Film one continuous three-minute shot of a place with activity around the camera — a workshop, a kitchen, a rehearsal. Ambisonic mic at the camera. No cuts. Export with metadata, watch it in a headset, and note every moment you wanted to look somewhere and couldn’t, or looked somewhere and found nothing.
That list is the whole craft.