[FLAW] Model cannot pronounciate seperate characters correctly

#26
by simzhou - opened

Description text:

casual random speaker, casual talk, male 30s.

Target text:

ECO mode

It should be pronounced as either 'E C O' seperately or /iː.kəʊ-/

Welcome to no phonemes TTS. Suffer!

Maya Research org

Normalization is something common between all TTS models. Use regex to improve that part, or maybe try reducing the temperature of the generation for more stable generation. Regex takes up hardly 5ms.

bharathkumarK changed discussion status to closed

Normalization is something common between all TTS models. Use regex to improve that part, or maybe try reducing the temperature of the generation for more stable generation. Regex takes up hardly 5ms.

Could you elaborate on this please? It is still unclear how regex should be used in that case. Thank you!

Sign up or log in to comment