Have you experienced issues with speech to text failing? Could we look into implementing some form of consensus among the available providers?

[ryhamz]

If text to speech isn't a bottle neck, then disregard this.

If speech to text produces an appreciable number of errors though, I or we could look into gathering a consensus value from the available tools.

e.g. If Amazon produces a string (s1), but Microsoft and Mozilla are in agreement on a string (s2), we can toss out Amazon's string and proceed with s2.

[mikeyy]

This is something I've thought over although I haven't got around to implementing certain measures to mitigate avoidable repeated failures, including fallback measures for incorrectly deciphered audio for each services. I'd also like to attempt storing parts of the audios metadata, in the form of an MD5 hash or the likes as some sort of cache, along with it's success and it's answer or vice-versa.

Unfortunately, I'm lacking funds to actually pursue this task anytime soon. It is my deepest regret that until then this project's future is at a full-stop.

[ryhamz]

Sounds good. I was actually coming at it form the angle of "would this be worthwhile to attempt building out myself in order to pull request in? It could be fun.

Totally understand that you see this project as near-ceased, in terms of your involvement.

[mikeyy]

Oh, well in that case, most definitely - go for it!