Open earboxer opened 6 years ago
Based on performance issues with mimic, (https://github.com/MycroftAI/mimic/issues/155 https://github.com/MycroftAI/mimic/issues/77), it seems like the best option is to create audio files of the voice, then combine them with the backing track using something like pydub.
These songs were hand-created by various users in audio editing software. I'm happy to pull in more songs if others want to create some, but at this point I don't expect Mimic or Mimic2 to be able to generate properly timed music without someone spending a lot of time on it. I can imaging using some of the karaoke tracking software to pull something off, but this isn't something the core team is going to exert effort on.
Currently, the audio in this skill is hard-coded.
Mycroft Mark II claims
To support these voices, this skill will either explode with more files, or need the text coded in and audio generated at runtime.
I suggest the latter.