Skip to content

YazSes v2.10.0-dev.5 — Wave N complete: Echo + SRPace (developer preview)

Pre-release · not published (publish workflows are pinned to v0.x/v1.x; v2 dev tags do not ship). v1.4.1 remains the stable release. Every feature is OFF by default.

Completes Wave N (design/vision/v2-research/16-wave-n.md) — structural editing, i18n & accessibility-output correctness — with the eyes-free & blind-output tier. All ten Wave N features are now shipped.

New in dev.5 (all off by default)

  • Echo (ADR-v2-119, [echo]) — "play that back" replays your own captured audio for a text span (not a TTS re-reading), so a homophone or ASR error is exposed by ear, eyes-free. build_span_index maps char ranges to audio time ranges from word timings; then resolve_playback_target turns "play the last word", "repeat 'foo'", or "play that back" into an audio (start, end) window. Anchor: Vertanen eyes-free ASR-error detection (arXiv 2410.20564).
  • SRPace (ADR-v2-120, [srpace]) — pace injected text to a screen reader's reading rate, clause-chunked, instead of dumping a paragraph as one burst that floods the reader. chunk_clauses + plan_injection_schedule assign each clause a cumulative delay from your configured words-per-minute so the screen reader announces each piece coherently. Anchor: screen-reader comprehensibility research.

yazses features now lists 135 capabilities.

Wave N complete

All ten features shipped (ADRs 115-124): HatSelect · Transliteration · BrailleOut · Spoken Outline · Diacritize · SafeGlyph · WordFind · LoadGuard · Echo · SRPace — structural code editing, internationalization, and accessibility-output correctness, every one pure, 100%-covered, and off by default.

Quality

1453 tests green across Linux × macOS × Windows on Python 3.11 and 3.12. Both new cores are unit-tested to 100% with no new dependency (pure span index + pure clause scheduler; audio playback and the injector's delay honouring are existing I/O). The default install and the v1 dictation path are unchanged.