I just thought of this idea. There’s no database containing the correct pronounciation of any artist name or track.
Google and Apple make up for this by just making it up on the spot.
It would be interesting if MusicBrainz could offer the data of correct pronounciation of the name of an arbitrary artist or recording in the database.
We’re severely behind regarding speech recognition in the FOSS community anyways, as far as I can tell. Maybe it would be fun to implement this. (Not that I’d know.)
For the artist name, I think Wikipedia often tells it with IPA syntax (international phonetic alphabet?).
For the track names, actually work names, they are often made up of existing words.
But I don’t see the link with speech recognition.
You would like editors being able to set artist name pronunciation by (correctly) speaking in their computer?
Editors may not pronounce the names correctly, depending on their origins.
Our do you mean text to speech / speech synthesis, once the pronunciation is known by MB, so that it can say it out loud to the user?
Sounds similar to Forvo - https://forvo.com/
Yeah, but that’s not good enough. Computers can’t pull off a pronounciation just from IPA.
But I am fantasizing.
That’s the whole point of this post. Every weird artist name has a weird pronounciation to go along. Speech recognition engines do not know what that is because nobody trains them on that data, as it would take too much effort and would potentially poison the rest of the training data.
Not that I’d know.
Regardless, if someone were to tackle this, where would this info be stored? Whether or not it is stored here, we would probably need to provide the UUIDs for each artist and recording title.
I don’t see this taking off. But you know that.
And just to push this idea to its obvious final conclusion:
Obviously, sometimes there’s no place to even begin. I know that all too well…