-
Notifications
You must be signed in to change notification settings - Fork 23
Get Started for NNSVS (WIP)
Before labelling a NNSVS database using the NNSVS Singer labeller, make sure your file directories are set up the following way:
- my_nnsvs_database
- wav
- 1.wav
- 2.wav
- lab
After opening vLabeler, go to Start > New Project...
to create a work environment.
Sample directory - directory of your singing database. Contrary to the name, you point it to the working directory instead of the wav
folder. (in this case: no7singing
)
Project name - the displayed name of your project.
Project location - location to which the project files will be saved.
Cache directory - location to which the cache files will be saved.
From Category, pick NNSVS
. The only supported Labeler from this category will be the NNSVS singer labeler
.
On the right of the drop-down list, there's a ⚙ icon. Hitting it opens a pop-up menu that lets you set the paths of your data folders within the working directory. wav folder name
could point to the wav
folder, or the wav_O
folder in this case (both of them contain different audio files). One wav folder can be edited at a time.
Create by... section lets you pick between two options:
Default
- creates an empty label sub-project for every audio file. <-- Preferred option
Template generator
- lets you generate a template sub-project for every audio file:
Audacity labels to Sinsy lab format - converts pre-existing Audacity-style labels to NNSVS-compatible label sub-projects
Lab generation with ust (ja-kana) - converts pre-existing USTs in Hiragana to label sub-projects
Lab generation with ust and custom dictionary - converts pre-existing USTs in a custom phonetic system to label sub-projects
Regex raw label generator - generates label sub-projects from a Regex command
Once everything is set up, hit the Finish button. vLabeler should proceed to the project editing mode, provided each path and setting had been entered correctly.
Navigate and search
Edit parameters
Scissors tool
Scroll and zoom
Audio playback
-
NNSVS Singer labeler
doesn't update the project file after adding new*.wav
files to yourwav
folder. You have to re-make a project file in order to edit the new files and optionally merge it with your older project file. Edited existing*.wav
files are updated, though.
NOTE: Output labels from this labeler can be used for DiffSinger databases. You can later convert NNSVS-style labels together with WAV files into a ready-to-train DiffSinger database using a dataset format converter.