NV Speech Player – Recent Update Highlights
This update focuses on voice quality, single-word clarity, and customization, with new engine-level controls and better tooling for voice authors.
New: Single-word speech smoothing (language pack setting)
A new group of language-level settings improves how single words or letters are spoken (for example during character echo, word-by-word navigation, or key feedback).
These settings:
- prevent abrupt cut-offs at the end of isolated words,
- avoid unintended rising “comma-like” intonation when reading a single word,
- and make standalone letters like R, are, and you sound complete and natural.
Languages can opt into this selectively, so UI phrases and normal sentence reading remain unaffected.
New: Voicing Tone support (engine + packs)
NV Speech Player now supports voicing tone control, allowing voices to differ not just by pitch or formants, but by glottal and spectral character.
Voice profiles can now define:
- glottal pulse shape,
- pre-emphasis balance,
- high-frequency shelf EQ,
- and overall spectral tilt (bright ↔ dark).
This makes it possible to create clearer distinctions between voices (for example adult, feminine, or child-like qualities) without duplicating phoneme tables.
Phoneme Editor updates
The phoneme editor has been updated to:
- edit voicing tone parameters, and
- apply them directly to custom voice profiles.
Voice balance improvements
- Updated default voice profiles (including Beth and Bobby) to better match overall loudness expectations. Still not great, but a little better. We must avoid over-gaining it else it'll squish at 11025.
Language pack refinements
- Fixed the standalone English letter “A”, which could previously sound awkward.
- Removed outdated normalization rules that interacted poorly with newer timing logic.
- Improved consistency for languages with phonemic vowel length or open vowels (for example Hungarian and Finnish).
- Minor tuning for Spanish trills and palatal sounds to better match native expectations.
Under the hood
-- First build to include Linux support alongside main binaries.
- NVDA driver's bugs fixed: We no longer delay you when switching away from SpeechPlayer (although, could have made that a 20 second delay you know, be thankful it was just 2 seconds), and crucially, we properly terminate the DLL's handle and thread when exiting, so no stale Speech Players' lying around. We also fixed the driver to keep your custom voice set even if you hit escape and cancel changing settings.
The driver continues to become more and more modular as code complexity increases.
Notes for testers and users
If something sounds off, please include:
- the exact word or phrase,
- the language or dialect,
- and whether it happens when reading a single word by itself or within a sentence.
This helps distinguish phoneme issues from timing or prosody behavior.