Accessible STEM / visual reference layer
Captions say what.
Deictic shows which.
Deictic resolves the visual meaning hidden inside words like “this”, “that”, and “here” — so a deaf STEM learner can follow the equation the lecturer is pointing at.
Built around one interaction: spoken reference → visual target → explicit caption → clickable highlight.
Core interaction00:30.0
Normal caption
“Now this becomes negative.”
visual referentresolved
−v₀equation term
One narrow barrierVisual references disappear when speech is captioned alone.
One concrete actionClick the reference and the exact target lights up.
One honest proofBaseline, resolved, and adversarial metrics are replayable.
For builders / bounded extensions
Ready to test another domain?
The extension lab adds energy-transfer fixtures, object targets, and a browser-local custom-video sandbox with explicit diagnostics.