Toki Pona is a minimalist constructed language, both in words (~137) and phonemes (14). Because of this and the relatively strict phonotactics, it has just 92 valid syllables. With the writing system being phonetic (it's literally just pronounced like in the IPA), I wondered if it would be possible to make a Toki Pona text to speech script that merely strings together pre-recorded syllables. Obviously this was never going to work very well, but it isn't the worst sounding TTS. With better samples this could actually sound pretty ok
NOTES: ---------------------------------------------------------------------- This does not support stress, but stress isn't phonemic in Toki Pona anyways so it's not that important I said earlier there are only 14 phonemes, each represented by one letter, however there is a bit of an exception: Yupekosi, meaning "change a creative work and unintentionally make it worse," has the letter ⟨y⟩. "No one knows how to pronounce the y," which poses a bit of a challenge. Some people pronounce it /j/, however for objectivity (I guess) it will be pronounced as it is in the IPA, /y/. This is like the English "ee" sound or Spanish "i" sound, but made with rounded lips. Also, ⟨n⟩ is a word which technically breaks Toki Pona's phonotactics, but it is a word used similarly to "um" so it's ok. This is supported. Syllables not allowed do to unique phonotactical rules, namely /ji jin ti tin wo won wu wun/ are still supported despite them being forbidden. CREDITS: ---------------------------------------------------------------------- Syllables were sourced from http://ipa-reader.xyz/ with the voice "Miguel [US Spanish]"
