Fòram na Gàidhlig

I'm creating a Scottish Gaelic dictionary for a popular vocal digital synthesis application (Synth V), so it can sing in Gaelic. To do this I need the Gaelic word, and its pronunciation as closely as possible in IPA or Vocaloid X-SAMPA. I note that the dictionary on this site provides near IPA for Gaelic words. Is it possible to obtain a word/IPA extract to integrate into this application? Thanks.

Hi Andrew. Just to check first, you do know there is a pretty good Gaelic voice?

That's text-to-speech, which is a much simpler challenge. What I'm using is a vocal speech synthesiser which reproduces singing from text/midi notes based on a sampled human singer. Today they can produce pretty realistic singing in a small number of languages, including English. They do this by breaking words into phonemes and then synthesising the phoneme samples from the human singer together. I can take the 44 standard phonemes for English and use them to reproduce reasonably well the sound of Gaelic words being sung - but I need a dictionary to tr*nsl*t* the written word to the spoken phoneme set. That's what I am seeking. Once complete anyone can use the Gaelic dictionary for the synthesiser to produce synthesised Gaelic singing.

Here's an example of an English song vocally synthesised in this way... https://m.youtube.com/watch?v=xEcLnZy1yCA

Well that's a pretty huge task, and pretty ambitious.

I'm sure you're aware of most of the technical challenges, but how much do you know of the language-specific ones? (Genuine question -- you write in English after all, so we don't know what your level of Gaelic is.)

The biggest hurdle is that it's only the headword form that is given in the dictionary entries, and while some inflected forms are simple affixes (for which you won't get the IPA from the dictionary, but there's a limited set of them), there are inflections and derivations that change the root, most commonly by consonant mutation at the start or end of the word, but sometimes by syncope (loss of syllable) which occasionally has other knock-on effects.
(You may already know this. Not wanting to patronise, just making things clear.)

And the biggest challenge, the “44 standard phonemes for English” are only standard for English. Gaelic has completely different phoneme inventory, it’s impossible just to map Gaelic phonemes into English and keep authentic Gaelic pronunciation (but I guess it depends on what you mean by “reasonably well”…).

I'm also somewhat dubious as to what the point is. Of course with STT and TSS there are issues of accessibility i.e. they're useful to Gaelic speakers who are blind/deaf/dyslexic etc.

But I fail to see how vocal synthesis would aid Gaelic speakers? If successful, wouldn't that mean that we lose yet another domain for which a command of Gaelic is essential? There are precious few of those left, singing being on of them.