Why is live translation in smart glasses delayed?
Translation may wait for enough speech to form a useful phrase before producing a result. Speech recognition, translation processing, network transport and delivery to the glasses can each add time. The result can also be revised when later words clarify the sentence.
Fast fragments and useful meaning are different measurements
A few early words may arrive quickly while the key detail appears much later. Waiting for more context can improve meaning but slow the exchange. A connection problem can add delay without changing the translation itself, so one total number cannot explain every cause.
Measure from the end of a spoken phrase to the first translation a listener can use, and separately note revisions that change its meaning. An early placeholder or half-sentence should not be counted as a completed translation.