Link to issue number:

Fixes #15739.

Summary of the issue:

When there are several consecutive utterances of speech such as during say all or spelling, the pauses between utterances gradually decrease over time.

Description of user facing changes

When using say all or commands which spell text, pauses between sentences or characters no longer gradually decrease over time.

Description of development approach

Previously, the WASAPI code accumulated the total duration of ms buffered so far. To do this, we converted from the number of frames sent to ms and added that to the total ms. This conversion isn't perfect, which meant we continually accumulated conversion errors. This resulted in the callbacks gradually getting ever more inaccurate (earlier) until the stream was stopped and reset.

Instead, we now accumulate the total number of frames sent. We still convert to ms, since the current playback position is time-based. However, we only convert to ms on demand, so we don't accumulate conversion errors.

Testing strategy:

Set synth to eSpeak, rate boost enabled, rate 40.

Spelled the following line of text:

Another way to detect if this issue will occur is, at the noted rate values I posted above, spell a

Before this change, the pauses between characters would get faster and faster. After this change, the pauses are consistent.

Known issues with pull request:

None.

Code Review Checklist:

• Documentation:
  • Change log entry
  • User Documentation
  • Developer / Technical Documentation
  • Context sensitive help for GUI changes
• Testing:
  • Unit tests
  • System (end to end) tests
  • Manual testing
• UX of all users considered:
  • Speech
  • Braille
  • Low Vision
  • Different web browsers
  • Localization in other languages / culture than English
• API is compatible with existing add-ons.
• Security precautions taken.