Train a Diffsinger model locally using Difftrainer tutorial 2026

Get Difftrainer: https://github.com/agentasteriski/DiffTrainer The most consistent problem with installing is requirements not being installed! This vibe coded script seemed to fix the issue: https://pastebin.com/Z8zv1iN5 Here's the phoneme extractor: https://maeblythe.neocities.org/getphonefromlab I mentioned dsdict.yamls can't be copyrighted. If someone included a whole dictionary, that could be copyrightable. But for a Japanese model? not unique enough! your model CANNOT SING without a dsdict.yaml!! Add the dsdict.yaml files to each folder inside the folders inside the bank inside the OpenUtau singers folder. Not every folder needs it, but it won't break anything to have extras.