The problem with SAPI4 voices is that there's no fixed range for the parameters, which means that different voices may have different minimum, maximum, and default values for rate, volume, pitch, etc.. This makes applying the same parameter across different voices difficult. When switching between voices, what should we preserve, the absolute value, or the relative percentage like in SAPI5?

For volume, we may want to preserve the relative percentage, so that 80% volume is always 80%, no matter the minimum and maximum volume.

For speed, the unit is words per minute. Should we preserve the words per minute, or preserve the relative percentage, because different voices may have different default speeds?

For pitch, the unit is hertz. We may want to preserve the relative percentage, because male and female voices have different default pitches. We don't want to make male voices use the same high pitch as female voices, or vice versa.

Another problem is that using relative percentage will change the meaning of the parameter values. For a male voice, a pitch of 40 might used to be too high, but now 50 means the default pitch, which may confuse users, and cause problems when upgrading.

I would like to hear your thoughts about this. As SAPI4 is being deprecated, we may also choose to not fix this for SAPI4.

cc @SaschaCowley @LeonarddeR @jcsteh

For SAPI4, I decided to preserve percentage for volume, and preserve the difference between current value and default value for rate and pitch, without changing the meaning of the parameter value.

Personally, I'm not convinced we should preserve the values at all, precisely because different voices can apply different values very differently. With a synth like eSpeak or even OneCore or Vocalizer, it's okay because the voices respond to parameter values fairly consistently; all the voices effectively use the same underlying engine by the same vendor. With SAPI4 and SAPI5, this is not true at all: two different voices could have two entirely different synthesis engines. Rate 100 with one voice could be twice as fast as some other voice. I realise users find the current behaviour surprising, but my feeling is that we're just going to replace one surprising behaviour with another and we'll get bugs like "the voice is unintelligibly fast when I switch voices with SAPI5".

#17693 reports a problem that the rate value may become out of sync with the actual rate when switching to another SAPI5 voice. That's because when implementing rate boost for SAPI5, the rate value is changed to be stored inside a variable, which isn't reset when calling _initTts.

If we should reset all parameters when switching between SAPI4/SAPI5 voices, I will make another PR that resets the rate value properly and turns off rate boost consistently when switching voices.

Thanks for your insight. I will wait to see what others think about this.