What Is a Voicebank in UTAU?

Updated

A voicebank is the set of files that gives an UTAU singer its voice. It has two essential parts: audio files, which contain the actual recorded sounds, and configuration files (most importantly oto.ini), which tell UTAU how to cut, time, and name those sounds so they can be matched to notes. If you are choosing a voicebank to sing with, or building one to release, understanding these two halves is what determines whether the result sounds on-tune or broken.

The two essential parts

Audio files: the raw voice data

Audio files are the data a voicebank uses in UTAU to sing. They are short recordings of individual sounds or sound combinations, saved as .wav files. The names of these files reflect what they contain, and they can use either romanized or Japanese naming. Examples from the site include:

  • n_ba_bi_bu_be_bo_bo.wav
  • みみむまめみめ.wav
  • a_tsu.wav
  • ご.wav

On their own, these files are just recordings. They only become a usable voice when the configuration files tell UTAU where each one starts, ends, and belongs.

Configuration files: oto.ini and friends

Configuration files — more officially oto.ini files — split, time, and name audio files for use in UTAU. They do two distinct jobs:

  • Some configurations tell the program where to begin and end the sample.
  • Others tell the program the "perfect" timing to match the recording with a note.

This is the layer that turns a folder of .wav files into something a UST can actually conduct.

Aliases and frequency files

Two more pieces matter once you start using or sharing a voicebank:

  • Aliases are names for cut samples that allow a UST to "command" a singer to summon that sample. In practice, the alias is the label UTAU looks up when a note calls for a particular sound.
  • Frequency files (.frq) serve as pitch information from a recording. They help UTAU "flatten" the pitch so the singer can sing on-tune.

Other parts of a voicebank

A voicebank also contains folders, a readme.txt, and a character.yaml. These are less about how the voice sounds and more important for managing and releasing voicebanks for different programs — for example, the character file carries the singer's identity and presentation details.

How the parts fit together

Part Role
Audio files (.wav) The recorded sounds the voicebank sings with
oto.ini Splits, times, and names the audio files
Aliases Names that let a UST command a specific sample
.frq files Pitch information that helps UTAU flatten pitch for tuning
Folders, readme.txt, character.yaml Management and release details for the voicebank

What this means for you

If you are choosing a voicebank, the audio files decide the tone and language coverage, while the oto.ini and .frq files decide how cleanly and on-tune it will sing. A voicebank with strong recordings but poor configuration will still sound wrong.

If you are making a voicebank, plan for both halves: record the samples, then invest the time in oto.ini so each sample is cut and timed correctly, and generate .frq files so pitch is handled. The readme.txt and character.yaml come later, when you prepare the voicebank for release.

For a broader walkthrough of how UTAU works — the "studio," the UST as sheet music, and the voicebank as the singer — see the overview on utauguides.carrd.co, which organizes material by skill level (Overview, Basic, Advanced).