I have nothing additional to add, but I support this feature request and could benefit directly from it.

Yes, at its simplest level this could be bolding a word (or part of a word) for emphasis, and ending a sentence with a question mark for raising tone.

The longer term implementation should be SSML which most competing products already have.

yeah, more prosody control would amazing. "Style" starts to get there, but it's hard to really match to the meaning of the writing.

That would be a great feature. Is there a way for the word to lower at the end of a sentence? Is there a current setting for that?

Good points: emphases and question intonation could be simulated, if only manually. I'd welcome this feature.

Hargitai, how are you using punctuation to achieve this--for example, in a sentence like "I wanted one, not one"?

This is the main missing feature

I agree, Hargitai. While Overdub automatically handles normal punctuation, such as pausing after a period, presently it does not appear to raise pitch when speaking a sentence ending in a question mark (i.e., it must raise pitch to indicate a question and prompt a response; instead, it appears to maintain pitch as if speaking a simple statement). NOT GOOD!

Concerning Overdub requiring more training, perhaps its engine could benefit from such. That said. it should already (today) raise inflection for a sentence ending with a "?".

To better empower Descript users, I am recommending Descript use Speech Synthesis Markup Language (SSML) tags to provide additional control over how speech is generated from text (SSML is, in fact, how Alexa, Google Assistant, Siri, etc control text processing for speech intonation, pausing, etc).

To that end, please see my request entitled "Provide additional control over how Overdub generates speech from the text (e.g., using SSML tags)" here: https://feedback.descript.com/feature-requests/p/provide-additional-control-over-how-overdub-generates-speech-from-the-text-eg-us