[Feature request] tts from phonemes instead of words?

[asiletto]

I have a text which contains a foreign word. Using the english model, the word is spoken as it would be english.

tts --model_name tts_models/en/vctk/vits --speaker_idx p273 --out_path test_gobekli_tepe.wav --text "At a casual glance, Gobekli Tepe looks like an ordinary hill." I could obviously use the corrispondent model just for the foreign part, but the voice would be different.

Ideally I'd like to do something like this:

tts --model_name tts_models/en/vctk/vits --speaker_idx p273 --out_path test_gobekli_tepe.wav --text "At a casual glance, <phonetic>ɟœbecˈli teˈpe</phonetic> looks like an ordinary hill."

Or even just be able to give in input just the phonemes (which I am not sure is correct, but I could approximate it with multiple tries)

tts --model_name tts_models/en/vctk/vits --speaker_idx p273 --out_path test_gobekli_tepe.wav --phoneme "ɟœbecˈli teˈpe" Or can someone suggest an alternative to get "Gobekli Tepe" pronunced correctly with an english model?

[WeberJulian]

Hey, two response to that.

First espeak as a phonemizer is supposed to do that for the words it knows, for exemple "deja vu" with language set to english, is gonna be replaced by French phonemes.

Second, I guess the solution for that would be SSML, we've stop working on that because it's not a priority for now, but you can check it out here: https://github.com/coqui-ai/TTS/pull/1452/files

[asiletto]

Thank you, I'll look at the pull 1452

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. You might also look our discussion channels.