Norwegian UTAU development log 01

Alright hinga dinga people, I've started work on a phonemizer for Norwegian that I'm intending to be used with OpenUtau. I've found these datasets for pronounciations by Språkbanken, which I can use to train G2P models. What the G2P is, is an AI model that is trained on words + pronounciation in order to take the lyrics put in UTAU and turn them into the correct phonemes. Even words it doesn't know, it will try to predict how it sounds!

Status so far: Up until this point I've just been reading documentation and trying to wrap my head around how phonemizers and g2p work. I'm starting to get it now though, and so I've started prepping for training! I'm currently setting up the phones.txt file, which will tell the model what's vowels (stretch) or now (don't stretch)

I'll have to see when I can do the actual training though, I fear my laptops may be too weak? Hopefully they can do it :) Otherwise I have my gaming PC at my dad's place

Here is the data I will be using from Sprakbanken's Github

EDIT: First draft of reclist complete (it'll probably need adjustments) and phones.txt as well. I realised I need to remove the stress indicators from the training data... Just gotta find an easy way to remove numbers from a text file