[Feature request] Return timestamps for TTS output

[AlexLavSi]

Hey, would you mind adding some sweet new features? It would be amazing if we could get timelines for sentences in the text, maybe as a JSON file or something.

[eginhard]

It's not something we're planning to do ourselves in the near term, but I'd merge PRs adding this. If someone wants to work on it, best to submit a rough plan for feedback here first to agree on a common structure that would work with all Coqui models.

[AlexLavSi]

Thank you for your response.
I hope that this improvement will be both needed and liked by someone and that it will be made quickly. I would like to propose this variant of the JSON file (based on my needs):

{ "duration": 500, "timestamps": [ { "text": "bla bla bla 1.", "start_time": 0.0, "end_time": 3.5 }, ... { "text": "bla bla bla end.", "start_time": 480.7, "end_time": 500 } ] }

And text parsing can be done by sentences in the main text.