Skip to content

YazSes v2.2.0-dev.2 — Wave F complete (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 F — the full 10-feature SoA round from design/vision/v2-research/08-wave-f.md (ADRs adr-v2-035..044). Adds the six features beyond the dev.1 ship-now tier:

New since dev.1 (all off by default)

  • Acoustic Context Profiles (ADR-v2-039, [acoustic_profiles]) — detect the environment (quiet/café/car/meeting) and auto-tune the mic gate + noise suppression, with hysteresis.
  • Mood Ledger (ADR-v2-040, [sentiment]) — private speech-sentiment journal; labels stay in the encrypted corpus.
  • Pronunciation Feedback (ADR-v2-041, [pronunciation]) — per-phoneme good/fair/poor practice scoring for L2/accent training.
  • Personal Read-Back Voice (ADR-v2-042, [tts] clone_voice) — proofread in a clone of your own voice; permissive OpenVoice V2 default, embedding in the encrypted corpus.
  • Gesture Chords (ADR-v2-043, [gesture]) — bind multi-input chords (held key + nod / key / sEMG squeeze) to actions.
  • Two-Way Live Interpreter (ADR-v2-044, [interpret]) — face-to-face mode; each turn is detected and translated into the other language, reusing the offline translate + TTS path.

Plus the four dev.1 ship-now features: Speaking Coach, Smart-Paste, Audio-Anchored Scrubbing, Dictation Reflow. yazses features now lists 55 capabilities.

Fixes

  • features: corrected a config-section collision so yazses features enable prosody (and predict / spoken-edit) writes to its own [section], not another feature's; added a write-target regression guard.

Quality

1086 tests green across Linux × macOS × Windows on Python 3.11 and 3.12. Every Wave F pure core is unit-tested to 100%; heavy models (scene tagger, SER, GOP, voice clone, sensors, S2S) stay lazy behind optional extras. The default install and the v1 dictation path are unchanged.