Placement & Localization

Every scene needs a starting point in the real world. How it gets one depends on the experience's tracking context: a world-tracked scene is placed, a geo-tracked one is localized against the map. Both are set per scene, with nothing selected.

Placement

For world-tracked scenes. Open Scene Settings › Placement.

Method

• Viewer Places The Scene – the default. A placement indicator follows the viewer's aim and they tap to confirm. The most flexible choice, and the one to keep unless you have a reason not to.
• Place Automatically – the scene appears by itself as soon as a suitable surface is found. No tap, no placement UI.
• Marker-Based – the scene appears on a printed image the viewer points their camera at, in the same real spot every time and in the same spot on every device that finds it.
• Start Without Placing – no placement step at all. The scene starts at the viewer's position. Choose this when a script positions the content itself.

Surface

For the two placing methods, the surface the scene targets.

• Surface – Horizontal Plane for floors and tables, Vertical Plane for walls. Only surfaces of this orientation are offered to the viewer or found automatically.
• Classification – narrow it further to a kind of surface: Floor, Ceiling, Table or Seat for horizontal, Wall, Window or Door for vertical. Any accepts whatever the orientation allows.

Placement moves the scene and turns it to face the surface. It never tilts it. Content you compose standing upright in the editor arrives standing upright, facing the viewer, whether the scene lands on a floor or on a wall. A painting authored facing you hangs facing you, and it hangs level however the viewer was holding their device when it was placed.

The two methods treat the classification differently:

Viewer Places The Scene	A requirement. The placement indicator only settles on a matching surface, which shows the viewer where the scene is meant to go.
Place Automatically	A preference. Once a matching surface has been found the scene waits for it and appears there when the viewer looks at it. Only when none turns up does it accept any surface of the chosen orientation, so the experience always starts.

Floor is the exception when placing automatically. It is the one surface every experience can rely on, so it is never held out for: the scene is placed ahead of the viewer at floor height as soon as a floor is found, and the lowest surface stands in where the kind can't be determined.

When the viewer places the scene the floor is a requirement like any other surface, because the indicator can't hold an experience up – the viewer can always point at the floor. Where the kind of surface can't be determined, the lowest one stands in there too.

A surface's kind is not guaranteed twice

Whether a device calls a chair a seat or a table can differ between one run and the next, and resetting tracking re-decides it. A scene tied to a specific kind of surface may therefore land somewhere else on a later run. If the scene only needs a horizontal surface, leaving Classification at Any avoids the question entirely.

A non-floor placement locates the floor first

Scenes keep a floor for physics and shadows even when their content sits on a wall or a table, so the floor is found before placement starts – one extra step for the viewer. If the scene doesn't need one, turn Floor mesh off in Scene Settings › Meshing & occlusion and placement begins straight away. Only a placement that lands on the floor itself skips the step, since it establishes the floor as it goes.

Window and Door fall back to Wall

Whether a device recognizes a door as a door depends on the door. A plain flush door in a plain wall is usually reported as part of that wall, and on iPhone and iPad there is no separate door classification at all. So both choices accept the wall behind them: on Apple Vision Pro a real door is still preferred where the room reports one, and everywhere else the scene places on the wall rather than waiting for a door that is never announced.

This matters most for Place Automatically, which works down the list in order and only settles for the wall once no door has turned up. When the viewer places the scene the indicator settles on either straight away, since it has no reason to make them wait for a distinction their room may never report.

Recognizing a surface needs a capable device

Telling a table from a floor from a wall needs an iPhone or iPad from 2018 or later (A12 chip or newer), and isn't available on the web. Where the kind of surface can't be determined, any surface of the chosen orientation is used.

Minimum Dimensions

For Place Automatically on a horizontal surface: how much room the scene needs, in metres across the surface.

A table smaller than this isn't used, and the scene is kept far enough from its edges to fit – so a glance at the near edge doesn't leave content hanging off it. On the floor, which has no meaningful edges, it sets how far ahead of the viewer the scene starts instead.

Leave it at zero when the scene has no particular space requirement. It doesn't apply to walls: a wall's measured size reflects how much of it the viewer has scanned rather than how big it is, so it isn't something to hold a scene against.

Markers

For Marker-Based only. Add the printed image the scene is anchored to, with an accurate physical width in centimetres, and a position and heading to offset the scene relative to the image.

Add more than one marker to re-localize across a larger space – they cycle in the coaching overlay.

Range limits

A marker works well near where it was scanned; as the viewer walks 50–100 m away the device struggles to remember it and content can drift. Use markers for anchored, in-place scenes rather than long walk-arounds.

See Choosing and printing a detection image for what makes an image recognizable – the same advice applies to markers.

Localization

For geo-tracked scenes. Open Scene Settings › Localization.

Method

• Automatic – visual localization (VPS) where it's available, which is the most precise option.
• Marker-Based – localize from a printed marker instead.
• Manual – the viewer positions themselves on a map and sets their heading.

Choose Marker-Based or Manual for places VPS doesn't cover: indoors, or somewhere the map data doesn't reach.

Visual Positioning Preference (VPS)

Shown for Automatic. By default Scenery picks whichever positioning system is available at the viewer's location. If one gives better results where your experience lives, name it:

• Automatic
• Apple (ARKit Location Anchors)
• Google (ARCore Geospatial API)

App Clips use Apple's positioning

The Google option is too large to fit in an instant App Clip, so choose Apple's if no-install delivery matters. See Publishing & Sharing.

Localization markers

Shown for Automatic and Marker-Based, because they serve a different purpose in each:

• Under Marker-Based the scene is anchored to them.
• Under Automatic they're a fallback – if visual localization can't place the viewer, the experience looks for a marker instead.

Give each an accurate physical width in centimetres, and a position and heading to offset it relative to your content.

Under Automatic the experience works down the options in order: Apple's positioning where the location supports it, then Google's, then a localization marker if the scene has one, and finally manual localization. Markers are the last rung before the viewer is asked to position themselves on a map, so adding one to an outdoor geo scene is worth the effort even when VPS covers the site.

What the viewer sees

A hint naming what to look for. While a scene is being placed or localized the viewer is told what to point at – the floor, a table, a wall, a marker. When automatic placement is holding out for a particular kind of surface and the viewer is looking at one that's too small, the hint says so rather than repeating the same instruction.

A way out of a marker hunt. A viewer who doesn't have the marker to hand can tap Place Manually (world-tracked) or Localize Manually (geo-tracked) and position the scene themselves instead of waiting for a marker that isn't coming.

A confirmation when the scene has its place. A brief message – Placement Complete for a scene that was placed, Localization Complete for one localized from a marker or a location. Turn it off with Enable Placement Confirmation in Scene Settings › General for a scene that should begin without any interruption.

Placement and anchors

Placement decides where the scene starts. Anchors decide where each object sits from there – so an object with a Horizontal Plane anchor still finds its own table, whatever the scene's placement method. The two work together: placement establishes the scene's origin, anchors position content relative to the world from there.

Written from source rather than a transcript – this documents a new editor section, so no transcript exists to cross-reference.