AudioLex trains the understanding of spoken words for the case where hearing and understanding come apart: the sound arrives, but it is no longer reliably decoded as speech. That happens after one-sided hearing loss, and a hearing aid alone does not fix it — it makes things louder, not clearer. The path from sound to word has to be practised again. Two modes, both aimed at the same thing: Learning mode — a word is played with its text visible. This builds the link between what arrives and what it means. Exam mode — the word is played, the card stays covered. You rate yourself. That rating decides when the word comes back: after a minute, after ten, after a day, a week or a month. 92 words and sentences are included: 72 in German — 68 from free speech synthesis, four spoken by real voices — and 20 in English. Which language you train is a setting of its own, separate from the language of the interface. Your own recordings can be added: speak a word or sentence, write it down, train it. Several speakers can be kept apart and switched on or off individually — a familiar voice is a different exercise from a stranger's. For harder practice, put a background noise under the speech and set the distance between the two in decibels. Three levels do it in one tap: easy without noise, hard with a clear speech advantage, advanced with the noise above the speech. One noise is included, recorded inside a bus. Your own can be recorded or imported as WAV files. For headphone training there is a channel choice — left, right or both — which helps when only one ear should be trained. Over a Bluetooth hearing aid, stereo is summed to mono; the app detects this and shows the channel choice as ineffective instead of pretending otherwise. What AudioLex does not do: it requests no internet permission, has no account, no trackers, no ads. Vocabulary, ratings and session history stay on the device — not as a promise, but because the app has no technical way to send anything. A backup