search

ultrasound1372 / NVDA-IBMTTS-Driver

This progect is to develop and maintain the NVDA IBMTTS driver. This repository is responsible for the x0_personal versions, which are continuing development even after it seems activity from the base repository has stopped.
GNU General Public License v2.0
1 stars 0 forks source link

Any chance of upsampling the Eloquence TTS audio output? #9

Closed amirsol81 closed 4 years ago

amirsol81 commented 4 years ago

This is more of a suggestion rather than an issue. If I'm not mistaken, IBMTTS uses 22000 HZ for its output. However, I'm wondering if we can upsample the output to something like 44100. Not sure if it can improve our listening experience, but it would be a good idea to give the higher sample a try. So any chance of doing it as a user-configurable option from NVDA's Synthesizer dialog?

amirsol81 commented 4 years ago

As a follow-up, Eloquence for Android has such an option and I think Nuance TALKS&Zooms used to have it. I'm not sure if their highest audio sample rate is higher than what we currently have, though.

Mohamed00 commented 4 years ago

Probably not. Eloquence has always ran at 11 kHz, and the upsampling is probably coming from SAPI, not the synth itself. Eloquence for Android has a 16 kHz option, but that never made it to the desktop or even Windows Mobile, you can only get it in Android or in Eloquence for Symbian. I'm pretty sure upsampling would probably not do much except cause weird interpolation artifacts, anyway.

On 7/20/2020 2:10 PM, amirsol81 wrote:

This is more of a suggestion rather than an issue. If I'm not mistaken, IBMTTS uses 22000 HZ for its output. However, I'm wondering if we can upsample the output to something like 44100. Not sure if it can improve our listening experience, but it would be a good idea to give the higher sample a try. So any chance of doing it as a user-configurable option from NVDA's Synthesizer dialog?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ultrasound1372/NVDA-IBMTTS-Driver/issues/9, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADY4AYBNSDDRHTA4GSIM44LR4SCAVANCNFSM4PCUGFJQ.

amirsol81 commented 4 years ago

Ah, I see. Just thought it could be achieved like some of the features made possible by this new development branch. Thanks.

On 7/21/2020 2:26 AM, Mohamed Al-Hajamy wrote:

Probably not. Eloquence has always ran at 11 kHz, and the upsampling is probably coming from SAPI, not the synth itself. Eloquence for Android has a 16 kHz option, but that never made it to the desktop or even Windows Mobile, you can only get it in Android or in Eloquence for Symbian. I'm pretty sure upsampling would probably not do much except cause weird interpolation artifacts, anyway.

On 7/20/2020 2:10 PM, amirsol81 wrote:

This is more of a suggestion rather than an issue. If I'm not mistaken, IBMTTS uses 22000 HZ for its output. However, I'm wondering if we can upsample the output to something like 44100. Not sure if it can improve our listening experience, but it would be a good idea to give the higher sample a try. So any chance of doing it as a user-configurable option from NVDA's Synthesizer dialog?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ultrasound1372/NVDA-IBMTTS-Driver/issues/9, or unsubscribe

https://github.com/notifications/unsubscribe-auth/ADY4AYBNSDDRHTA4GSIM44LR4SCAVANCNFSM4PCUGFJQ.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/ultrasound1372/NVDA-IBMTTS-Driver/issues/9#issuecomment-661356137, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIGLZMXQLC7WU2WPDOGDCHLR4S4P5ANCNFSM4PCUGFJQ.

ultrasound1372 commented 4 years ago

Nah, upsampling properly would add no more than there already is. And upsampling poorly, like SAPI does, would make it sound like ass, with zero-order holds causing extra highs to appear where there were none. Besides, I am not upsampling anything in pure python in real-time.