There are quite a lot of really, really large utterance/phoneme datasets out there, but I can't seem to find anything small. I'm looking for the minimum viable set of utterances/sentences in English that would provide sufficient coverage of phonemes so that, if you fed them all into a training algorithm, you could get a half-decent model. I don't actually want any audio data, because I'm going to generate that myself (custom voice model).
Someone shared: https://github.com/rhasspy/tts-prompts
This looks great!