Welsh Besedilo v govor

O projektu Welsh besedilo govora

Welsh text-to-speech has to master initial consonant mutations, a defining Celtic feature where the first sound of a word changes — cath 'cat' can become gath, nghath or chath — depending on the preceding word, so the engine must read mutated forms naturally even though they look like different words from their dictionary entries. Welsh orthography is famously consistent once you know it: letters like ll (a voiceless lateral fricative) and dd (a voiced 'th') are single sounds written as digraphs, and the voice must treat them as units rather than reading the two letters apart, while also handling the long words that consistent spelling produces. Stress in Welsh falls predictably on the penultimate syllable of most words, which gives the language a recognizable rhythm a good synthetic voice has to preserve to sound Welsh rather than English-with-Welsh-words. Demand here is unusually concrete: Welsh is an official language of Wales with a government strategy targeting a million speakers by 2050, driving real text-to-speech need in bilingual education, public-sector services, broadcasting, and assistive technology — including synthetic Welsh voices for people who lose their natural speech.