See-through mode sets the ceiling on real–virtual fusion

What it is

The same request — “put a virtual cup on the real table” — is not one notch better or worse on optical versus video see-through. The fusion that can even exist is a different kind. See-through mode decides whether reality enters the eye as photons or as pixels; alignment, brightness matching, and latency budgets all run under a ceiling that choice draws. No amount of content polish crosses an operation the pathway cannot perform.

The ceiling is architectural, not a finish. Treating “it looks fused” as something texture work can pile up locks the product to a promise the pipeline will never keep.

Why it happens

The two pathways operate on different objects. In optical see-through, the real layer is ambient light and the virtual layer is added light; fusion is “stack something on photons already there,” so photometric match, shadows, and dark regions inherit that additive constraint. In video see-through, the real layer is already an image in a buffer and the virtual layer is another layer in the same buffer; fusion is “compute in one pixel domain” — brightness can be graded together, compositing order can change — but reality no longer arrives as live photons, and delay plus reconstruction error become the new ceiling.

Milgram’s mixed-reality continuum writes this as where the display inserts itself between real and virtual. Different insertion points reach different places on the continuum. A common product failure is to advertise from one end and demand fusion operations that only exist at the other.

Studying it

Write down the fusion operations the product must make true (un-delayed reality, seating a virtual object on a table edge, joint darkening of both layers). Mark each as possible in principle on an optical path, a video path, or neither. Run the same content on two see-through devices — or two modes of one device — and collect coherence / consistency ratings (environment-consistency subscales of presence inventories, or a dedicated real–virtual consistency score), not preference alone.

Independent variables: see-through pathway, the required fusion-operation set. Dependent variables: whether an operation is possible in principle, consistency scores, the share of trials judged as “it is sitting on that table.”

Keep “which headset do you like” separate from “which pathway can carry this fusion list.” Mass, field of view, and brand drag the former.

Where it stops holding

A headset that switches between optical and video see-through has a ceiling that moves with the current mode; do not write it into a product promise as a constant. Optical schemes with light valves or tunable transmittance will shift the ceiling; evaluation must bind to the pathway under test, not to the category name “AR.” Remote collaboration and fully virtual levels do not fuse with the current room, so see-through mode stops being a fusion ceiling and is only a display means. With one viewer and nothing that must nail to a physical surface, the constraint barely fires.

Applying it

• At kickoff, write fusion operations as a checkable list: which must hold in the photon layer, which may hold in the pixel layer. Choose see-through mode against the list, not against a more-AR or more-MR label.
• If the list asks for an operation the current pathway cannot give, change the task or change the pathway. Do not let a composited trailer stand in for runtime capability.
• Promise only fusion the chosen pathway can repeat. Anything that must still hold after a venue or lighting change belongs in acceptance, not in a concept film.
• How to check: walk the list on the target headset — possible is a pass, impossible is an architectural gap. Ask someone new to the project, live, whether the cup is on the table or stuck to the glass. Split answers mean fusion has already climbed past this pathway’s ceiling.

Related

• Same group：N5.01.1 Optical see-through keeps real light but contrast is limited · N5.01.2 Video see-through is controllable but adds latency and image loss
• Nearby：N5.08 Video See-through and Optical See-through · N5.02 Virtual–Real Occlusion
• Search terms：see-through mode · mixed reality continuum · photometric consistency