Uploaded August 2023 | Updated September 2026, 1 hour ago

---

Like how this sounds? Check out this video by @hpsynthv7455 that explains what's going on here, with installation instructions in a video in the description!
youtu.be/jrAOAuJxIfU

If you want to go straight to the project, I was almost right about the name!
github.com/schufo/lyrics-aligner/tree/main

Song is Big Iron by Marty Robbins. This method requires isolated vocals as a reference for the AI!

I use an edited version of Hataori's textgrid to UST script that includes the intensity of notes for the dynamics and Utalis for the pitch bends.
shinta.coresv.com/software/utalis_jpn

Unrelated, but if the OTOs seem off, I actually wrote a little webpage that turns utan banks into utau banks. It seemed to work pretty well half of the time 😛 if I remember correctly, it is actually full diphone coverage except VV because of how English works and CC that would basically just be silence and make gross hissing sounds.

Your bank can use any phoneme system you want! If it's not arpabet, though, you'll need to convert the phonemes like I did 😝