[lucb1e]

For an IPA-to-sound converter, if the chosen voice doesn't have the sound that the user is asking it to render, it should: either throw an error, find an English speaker that can, or synthesise the sound altogether.

Giving you flat-out the wrong sound (not an accented rendition) should not be default behavior. There could be an option for rendering "how would an american approximate/recreate this sound after I sounded it out to them", but that's a different tool and not an IPA-to-sound converter.