ADR-002: Embedded audio announcements

Status

Accepted

Date

2026-08-12

Context

System speech synthesis was absent or inconsistent on the target display devices. A first embedded eSpeak asset set produced unintelligible Mandarin. The repository already contained a prior Mandarin queue-announcement WAV set that the user confirmed was the intended project audio.

Decision

Package the project’s legacy Mandarin prompt, digit, range, and suffix fragments in the APK. Parse and validate PCM WAV assets, compose the number or range locally, and stream each announcement three times using AudioTrack with media usage and transient audio focus. Do not announce the project name. Do not depend on system TTS.

Rationale

Deterministic bundled PCM avoids browser autoplay and installed-voice dependencies while allowing the server to continue sending ordinary snapshot text and batch data.

Consequences

The APK contains and tests exactly 13 WAV resources.
The player must own FIFO ordering, cancellation, focus, and native track lifecycle safely.
The original generator, speaker, and separate redistribution license for the historical WAVs are not recorded; confirm ownership before distribution outside the project's authorized deployment scope.
Audio routing and intelligibility remain target-device acceptance items.

Supersedes

Android system TTS announcements.
The rejected eSpeak-generated Mandarin assets.

android/app/src/main/java/cn/nianxx/queue/display/QueueAnnouncementPlayer.java
android/app/src/main/java/cn/nianxx/queue/display/EmbeddedWavComposer.java
.project-docs/50-evidence/topics/20260811-native-display-f2456ea9__embedded-audio-assets.md

1.7 KiB Raw Blame History Unescape Escape