In my peculiar little idiolect the set of skins of which she has "a" skin is not the set of skins she personally possesses but the set of skins which the human race as a group has available to it; of which she has a fair one. I apologise for cluttering this space with less relevant remarks but I feel the need to make that clear.
There’s no need to apologize. It reads as though you treat “skin” in the same way as “nose”, e.g. “He has a sensitive nose”, whereas I treat “skin” in the same way as “vision”, e.g. “It has excellent vision”. Perhaps the difference could be explained by feature (with an “a”) vs. characteristic (without an “a”)?
I know, but in fairness, I find the Text-to-Speech quite poor. For some reason, the Corca Dhuibhne is worse than the others, with quite often some artefacts that prevent from truly distinguishing the sounds. What I just discovered is the automatic phonetic transcription, but it only gives Gaoth Dobhair and Conamara. Or were you talking of something else?
I wasn't aware of this limitation with Duolingo, thanks for this clarification too!
abair.ie uses two different techniques for it's different "voices". The Gweedore and the first Connacht voices use actual recorded speech samples that are cut up and stitched back together. It seems to work a lot better for the Connacht engine than the Gweedore engine, which is much "choppier" in my experience. The Connacht HTS and Dingle HTS engines essentially use a statisical model the required sounds, and are supposed to generate the actual tones, rather than select them from a database of recordings.
I don't know why the HTS recordings are so "lo-fi" that they sound like the speaker is talking through a paper bag - I would have expected that the generated tones could be "purer", but it may be that they are based on an analysis of the speech of a speaker with a slightly hoarse or gravelly voice. It's a shame, because it makes them a lot less useful for hearing the finer points of the pronunciation.
The HTS engines seem to go offline fairly frequently too - the Gweedore and Connacht voices almost always available, but the HTS voices are often "silent", and when the voices are silent, the transcription is too, because, I think, the spoken sample relies on the underlying transcription.
The Corcha Dhuibhne transcription for orthu is 1 o . 0 ɾ̥ˠ ə and the transcription for uirthi is 1 i ɾˠ . 0 h ə.
Note that the Connacht HTS transcription for orthu (1 o . 0 ɾ̥ˠ ə) is not the same as the Connacht HTS transcription for uirthi (1 o ɾˠ . 0 h ə)