piper
piper copied to clipboard
Feature Request - Add Timestamp Functionality
Could we incorporate timestamps in the outputs? This is need for highlighting text while speaking.
I'm working on this, almost ready. Still has a few minor issues to solve. Waiting for another pull request to be merged, after this I will be hopefully done with it.
Is it the same issue as I asked about in discussion about Calibre? https://github.com/rhasspy/piper/discussions/391 That would be awesome.
And it looks like this is about the same issue: https://github.com/rhasspy/piper/issues/361
Originally I wanted to use a different output, but I might use this as an output format instead. I need to look into the format first, if that matches my needs. I need an output file with exact timestamps instead of continuous streams. Also, as mentioned in those tickets, using the engine word by word creates different output, so I run it twice. Once for audio and once for timestamps. Fine for "offline" use, but takes 2x the CPU time. Might still be good enough on powerful hardware.