NVDA Alpha 35286 finally adds leading silence detection and removal logic. The changelog says:
"The silence at the beginning of speech will now be trimmed when using OneCore voices, SAPI5 voices, and some third-party voice add-ons to improve their responsiveness."
Though not mentioned in the changelog, it also affects eSpeak NG.
NVDA 2025.1 is shaping up to be a huge release.
https://github.com/nvaccess/nvda/issues/17614
@NVAccess
@amir @asael @NVAccess The average user isn't going to reconfigure the silence threshold when their synth isn't pronouncing something correctly. I think if it were going to be configurable, it'd need to be disabled by default. But hopefully there's a better default threshold that can be used which would avoid this.
@jcsteh @amir @asael Thanks everyone for all the great feedback on this (note that GitHub is the best place for feedback, but we won't ignore feedback here if we see it (noting that in some cases - such as this - NV Access may not be the author of a PR - so do keep reporting things on GitHub - in any case, this one is being worked on, so do keep on the lookout for improvements in future builds. Thanks again!