Minimum viable English phoneme coverage set, LJSpeech-style?

[thoraxe (Erik)]

I’m looking at Mozilla TTS and taking the following into consideration:

Clear process for generating custom voice

Hi, Newbie here, so apologies if I’m missing the obvious. I am trying to achieve the following and think the steps are as follows, but have a few gaps. Please can someone help me fill in the gaps/answer the inline questions in Italics? And of course suggest any steps that may be missing. Maybe this can become a clear quick-start guide for this particular aim. Thank you. Aim: To install Mozilla TTS on a Linux machine, and fine-tune a pre-trained LJSpeech with a new voice of my own. Steps: 1)I…

Dataset · mozilla/TTS Wiki

How do I get started training a custom voice model with Mozilla TTS on Ubuntu 20.04?

TTS/notebooks/dataset_analysis at master · mozilla/TTS

Unfortunately, what seems to be to be a simple question does not seem to have an answer that I can find:

What is the minimum viable set of sentences/utterances in English that provides good phoneme coverage for generating a model with a custom voice?

There are several examples where people tried to do things and didn’t get great results, and no phoneme sets or LJSpeech-type data is available. There is the original LJSpeech dataset which is immense and is not organized in any priority-type order. There are other LJSpeech datasets that are still really quite large.

I understand English is a complex language. You would think, though, that it would be possible to find a minimum-viable dataset that covers all phonemes adequately for making recordings.

Does such a thing exist, and I’m just bad at searching?

[dkreutz (Dominik)]

This here claims that phrases are “phonetically balanced”, the “en-us” section contains 1.132 sentences, not sure if this is sufficient for a good dataset but at least it will be a good starting point:

GitHub - rhasspy/tts-prompts: Phonetically balanced text to speech sentences

[thoraxe (Erik)]

This is a great start, thanks!