Hi everybody, I'm making a database of VBs which have used any of my reclists, to be implemented on my website. If you've created a voicebank using any of my lists and it has a public demo (and preferably also a public download), feel free to submit it here:
Voicebank MUST have a public audio demo, and preferably a public download link (though unreleased and private banks are fine to submit). Sub
If you're unfamiliar, I made all the ones on this site:
English UTAU recommendation Doc (aka whenever people ask me for English UTAUs I forget every one I’ve ever used and this will help me not do that) Feel free to leave suggestions up here or at the bottom of the doc! Ms. Atia, Gekiyaku, Reine, and Lt would be here if they were public Type D...
Might as well put this here too-I made a document full of English UTAUs that I’ve personally used and can say are great quality! They’re sorted by type(CVVC, VCCV, Arpasing, etc) and come with notes on how to use them effectively. Leave a comment if you have any suggestions/corrections to make!
Also if you can find a good sample of Sayu Yurika singing in English that’d really help :’)
Parallel World, a plugin that lets UTAU be usable directly in a DAW, is now open to the public. Synthesis is done in real time, no need to wait for the audio to render!
Hello I’ve been meaning to post this for a long time, as someone who puts care into creating my UTAUs. I thought it might be useful to some people.
This is part guide and part inside look at what I do while creating my own UTAUs. I hope it helps someone.
Beginning Phase
First, I made a document using ideas I wanted to include. I design UTAUs based on tropes and qualities I like and want to see more often in characters. Here’s the list from 2019 of some qualities I wanted to give Lycoris. Some are subjective while others are specific.
monster qualities(three eyes/horns)
intimidating looking
bob haircut
modern clothing
Unique but not super dressed up
square jaw
broad shoulders
Thinking of what you want to see in an UTAU can help with the design process. Find different ways to put those elements together or add new ones!
Naming Conventions
It is extremely common for UTAUs to have Japanese names which are filled with meaning. Using the “Ne” kanji (which can also be read as oto) was common from 2008-2012 and is still used sometimes today. Using “-loid” or “-poid” suffixes is common, but may make your UTAU seem more like a software than a character.
I chose to give all my UTAUs non-Japanese names in order to make it easier on myself. Japanese naming conventions can be difficult to learn if you’re a beginner. Lycoris is a Greek name, Omichron is a made up name that’s Greek-ish, and Alta Maha’s name was inspired by the Altamaha-ha, a creature with roots in Indigenous Muscogee tradition. From the start I wanted all my UTAUs to have lore, and for them to be mythical/fictional beings of some kind. Making all your UTAUs in the same art style and theme can be fun, but don’t feel obligated to. Having a single UTAU with many updates and appends is also common.
Speaking of appends, the most straightforward way to name them is by voice type (Solid, Soft, etc) but unique names are fun too! I named all of Lycoris’ vbs after phases of the moon and have stuck with that choice.
A good tip is to not give your UTAU a name that already exists in UTAU land. Looking at Vocadb and UTAU wiki is an easy way to check for that. UTAUs are all part of the same “category” of vocal synths, so having a name similar to another can make it hard for people to find info about your UTAU.
Designs
Choosing to not give your UTAU a design is feasible, but that’s also a conscious choice. UTAUs may get more publicity than your OCs that aren’t tied to some other known property, so many want to give them good and memorable designs. Character design is subjective and a very complex topic, so I’ll just sum up some tips for UTAUs that I kept in mind.
The best advice for designs is to not trace any other art, official or not! If you trace, say, a Kei art, people will notice and their view of your UTAU may sour. Even using a picrew or a free to use base is better than tracing(as long as it’s allowed). In fact, there’s quite a few UTAUs in existnce who started out as picrew OCs. Commissioning another artist may be an option if money is no object.
Many UTAU designs were/are inspired by existing Vocaloids or the Macne series. The best advice I can give for these is to add some defining feature that makes them unique from their inspiration. There isn’t anything wrong with being inspired by an existing vocal synth as long as it’s not a blatant ripoff. My Omichron was loosely inspired by Macne Papa, but you probably wouldn’t guess that immediately!
I went through many, many concepts for Lycoris before settling on my final one. One of these eventually became the basis for Alta’s design! Other times, I find a design I want to settle on quickly.
This is the Lycoris concept that eventually became Alta. I largely was inspired by irl fashion.
I personally start with color palettes before drawing out an actual design, and when I settle on a design I adjust the colors further until they’re exactly how I want them. Colourpod is a good place to find palettes. Sticking to a limited set of colors means I’m not struggling to find some that look good together.
Creating a Pinterest board for your UTAU’s overall aesthetic can help to take inspiration from! It constantly gives new suggestions and ideas base on the images you add to it.
Standing Art
Deciding on a pose for your standing art can be tricky. My general tips here are to plan it in a way that shows off most of the design. Perspective shots are great but not for default standing pictures which may be reused for covers/other material.
You don’t have to make every standing picture perfect...I tend to do that and it makes them take forever, haha. Sketchy/simple official art can be just as pretty. And there’s all the time in the world to re-draw it.
Those are just a few tips for how I handle UTAU character designs! Nothing is a hard and fast rule of course. The best advice to creating UTAUs is to have fun with it.
UTAU - Generate all frequencies for a voicebank at once
This is a little-known feature of UTAU that I use all the time! There is a hidden button in voice bank settings that generates frequencies for every sample at once. It does take some time, especially for multipitch voices, but it’s easier than manually making each frq file.
This screen is located under Tools>Voice Bank Settings.
The hidden button is in between “Edit freq. map” and “Initialize freq. map.”
Double clicking the area between these buttons will make a pop-up appear.
This part of the interface isn’t translated but all of the three selectable options more or less do the same thing i.e generate frequency files for everything in the voicebank.
The upper button on the right starts the process. Generating will take some time, and make sure to press the Enter key when it's done to close the window.
I use this mostly for my own UTAUs, but also for some I download that don’t come with frequencies.
Selecting one folder at a time will let you generate frequencies pitch by pitch.
Most of the time this feature skips over some samples for some reason, unsure how to fix that. Still, it reduces rendering time a lot to just generate all the frqs at once.
I assume people who follow me here use UTAU, but I’ve noticed many people who want to get into UTAU but just...can’t.
UTAU is known as one of the hardest vocal synthesizers to get into, despite being free. Voicebank types are one of the first things you’ll come across when making them or downloading them. I answered an ask about it a while ago, but some might like a more deep explanation.
UTAU uses recordings of phonemes(sounds that make up a language) from a person in order to work. Finding a good recording list (reclist) is imperative before you record a voicebank. Generally, they contain all the phonetic data needed to put back together into a virtual singer. Making your own reclist is an option, but make sure it isn’t missing anything you need for your target language. Pronunciation is very important!
Speaking of languages, UTAU can only process Japanese hiragana and English letters. So if you download a Korean UTAU, it will not be in Hangul.
You can start recording with any language- UTAU users have made reclists for languages from Cantonese to Cherokee. Google will often point you to them but here’s some places I like to check.
UTAU Language Resources site
Utaforum Reclists resource
Misc. Tips
PC UTAU uses .wav files and Mac UTAU-Synth uses .aiff files. Make sure what you record is in one of those formats.
Any audio recording program will work for UTAU, Audacity is commonly used since it has noise reduction.
OREMO is a standalone program which was made for recording UTAUs-you load a reclist in and can check your progress. Akorin is an alternative if Oremo is too confusing or glitchy on your device.
When recording, sing your sounds and hold them out for at least 1 second. Also make sure to leave some noiseless space at the beginning and end of the recording-you don’t want to cut off the sample!
Voicebank formats
This is the aspect of recording talked about the most, but it is a technical topic. Remember, there is no such thing as a “bad” voicebank type - only an incomplete one. These terms also apply regardless of language, though the Japanese community labels some of them differently. I’ll also link downloads to an example of some of the reclists, but don’t feel pressured to use it.
CV/単独音 - CV stands for "Consonant-Vowel". The Japanese UTAU community uses the term Tandoku-on(単独音) for this voicebank type. Sounds consist of either a vowel sound or "consonant-vowel" (ka, ji, no, etc). It’is the earliest recording type developed for UTAU, and takes the least amount of time. It’s recommended for beginners due to its small size.
VCV/連続音 - Stands for "Vowel-Consonant-Vowel". The Japanese UTAU community uses the term Renzoku-on(連続音) for this type. It involves singing strings of at least a few syllables and then splitting them up through configuration. The result is often much smoother than what can be achieved with CV. In linguistic terms, VCV uses triphones in order to work. The average VCV reclist for Japanese is 190 samples long. Some new UTAU users are tripped up by the differences is reclist length and type, but VCV remains popular since it forces smoothness and clean results.
CVVC - Stands for “Consonant-Vowel/Vowel-Consonant”. Uses CV sounds, vowel sounds, and VCs (Vowel-Consonants) such as "ak" and "ek". It’s longer than CV but much shorter than VCV on average. It’s used often for many different languages due to being flexible.
VCCV - uses VC and CV sounds in a similar way to CVVC, but is far more expansive. VCCV was also developed specifically for American English: individuals have adapted it to other languages.
ARPAsing - A style made for English recording that uses the ARPAbet phonetic system. Developed by Kanru Hua, creator of Moresampler and Synthesizer V. This style involves singing words instead of sounds. It also doesn’t take many recordings to make, giving you a functional English UTAU within hours.
Others?
There are more voicebank types than the ones I listed! I just described the most common/popular ones. CV-C is one example of a reclist that can be used for a number of different languages. Some other variants have been created for different languages, like CVCC for English. There’s also modified reclists, like a CV with extras vowels or consonants. However, if you’re brand new to recording, it might be better to hold off on these since they require more knowledge of configuration, phonemes and OTOing. Speaking of OTOing...I can get to that later!
UTAU VOICEBANKS I decided to make an info-book for people who want to try out other UTAUs! (This was heavenly inspired by a post on Utaforum and a post on twitter by revolocities, except it’s without any subjective info). I take notes based on the criteria below. All of the notes that I took ...
This is a really cool doc! It was made by fwoofjay on Twitter and has a ton of popular and not-so popular UTAUs on it!
