简体中文
English
Website review
This review may include AI-generated content and is not individually verified. Domain, certificate and other facts come from the observed data on this page.
What is UTAU?
UTAU is a free, community-made singing synthesizer: you program notes on a piano roll, and a recorded voicebank "sings" them back as a rendered vocal track. The site frames the whole workflow as a collaboration with a real singer — you set up the studio (the UTAU program), write the sheet music (a UST), bring in the singer (a voicebank), then hit record and let the computer render the vocals.
The three pieces you actually work with
- The program — your studio. Everything you edit happens here.
- A voicebank — your singer. It is a folder of recorded audio plus configuration files (
oto.ini) that tell UTAU where each sample starts, ends, and how it should line up with a note. Aliases let a UST "command" a specific sample, and.frqfrequency files supply pitch data so UTAU can flatten the recording and keep it on tune. - A UST — your sheet music. It stores BPM, notes, aliases, pitch bends and vibrato, which is what makes a cover sound human rather than robotic.
Who it suits
UTAU appeals to people who want to make vocal covers, write original songs with synthesized vocals, or build and release their own voicebank — including character design, since voicebanks are often tied to an original character. The trade-off is hands-on work: unlike commercial synthesizers with polished defaults, results depend on how well you tune the UST and how cleanly the voicebank was recorded and configured. Expect a learning curve before anything sounds natural.
A practical first step
Pick the path that matches your level on utauguides — overview if the concepts are still fuzzy, basic if you know what UTAU is but not how to use it, advanced if you want to improve. Beginners should start with the overview pages on how UTAU works, what a voicebank is, and how USTs work before touching any settings.
For a concrete scenario: someone who wants to cover an existing Japanese song should download a UST for that song, load a voicebank, and spend their time on pitch bends and vibrato rather than on recording anything themselves. Someone who wants their own character's voice has the opposite job — recording samples, cutting them, and writing oto.ini entries — and should treat the UST side as the easy part.
How do I create a voicebank for UTAU?
Creating a UTAU voicebank means recording a set of sounds and then labeling them so the UTAU program knows when and how to play each one. The page frames it as giving a "singer" the ability to read a UST: the audio files are the voice, and the configuration files are the instructions for using it.
What actually goes into a voicebank
- Audio samples (.wav): the recorded syllables and phrases, such as
n_ba_bi_bu_be_bo_bo.wav,みみむまめみめ.wav, ora_tsu.wav. These are the raw material the singer uses. - Configuration files (oto.ini): these split, time, and name the audio files. Some entries tell UTAU where a sample begins and ends; others set the timing that lines the recording up with a note.
- Aliases: the names attached to cut samples so a UST can "command" the singer to summon that specific sample.
- Frequency files (.frq): pitch information taken from the recording, which helps UTAU flatten the pitch so the singer stays on tune.
- Supporting files: folders, a
readme.txt, andcharacter.yaml. The page notes these matter more for managing and releasing voicebanks across different programs than for the singing itself.
A practical order of work
- Decide what kind of voicebank you want and which sounds it must cover.
- Record clean, consistent samples, keeping the recording setup and tone as even as possible.
- Cut and name the samples, then build the
oto.iniso each one is split, timed, and aliased correctly. - Generate frequency files so pitch behaves.
- Test by loading the voicebank into a simple UST and listening for timing, tuning, and missing samples.
- Package the folder with its readme and character file if you plan to share it.
Where the real difficulty sits
Recording is the visible part, but the oto.ini is where most beginner voicebanks succeed or fail. A sample can sound perfect on its own and still sing badly if its start, end, and timing are off, or if its alias does not match what the UST asks for. Expect the labeling and testing stage to take longer than the recording stage.
If you are new to the concepts, the site's overview section explains how UTAU, voicebanks, and USTs fit together before you commit to recording; its basic and advanced sections are aimed at people who already understand those concepts and want to improve. utauguides
As a next step, pick one short, simple song, build a minimal voicebank for it, and render it. A small finished test tells you more about your labeling than any amount of reading.
What are the differences between basic and advanced UTAU techniques?
The site splits its guidance into three levels—Overview, Basic and Advanced—and the practical difference is not difficulty for its own sake but how much of the singing you control yourself. At the basic level you work with what a voicebank gives you: you load a UST, choose a singer, adjust obvious things like tempo and note placement, and render. At the advanced level you start reshaping the voice itself—tuning pitchbends, vibrato, aliases and timing so the singer phrases a line the way you intend.
What changes between the levels
| Aspect | Basic | Advanced |
|---|---|---|
| Main goal | Get a clean, on-pitch cover | Make the vocal expressive and natural |
| What you edit | UST as delivered, plus simple fixes | Individual notes, pitchbends, vibrato, aliases |
| Voicebank use | Pick one that fits the song | Match samples and aliases to specific notes and languages |
| Likely problems | Wrong aliases, off-tempo notes | Stiff or robotic phrasing, clashing consonants |
| Where time goes | Setup and rendering | Repeated listening and small corrections |
The site's own explanation frames UTAU as a collaboration with a singer: the program is the studio, the UST is the sheet music, the voicebank is the singer, and rendering is the recording. That metaphor is a good guide to the levels. Beginner work is choosing sheet music and a singer; advanced work is directing the performance.
A concrete example: a beginner renders a UST, notices one line sounds flat, and moves on. An advanced user hears the same line, opens the pitchbend curve, lengthens a vowel slightly, softens the vibrato on the held note, and checks whether the alias being triggered is the right sample for that syllable. Same song, very different amount of intervention.
Deciding where you belong
If you understand what a voicebank and a UST are but have not finished a full cover, start at the basic tier and expect the advanced material to be confusing without that experience. If you can already render a cover and your complaint is that it sounds mechanical, the advanced material is the right next step—phrasing, pitchbends and alias choices are usually where that stiffness lives.
For background on the software and voicebanks themselves, the official UTAU pages and voicebank distribution sites are worth checking alongside a guide like utauguides.
How do I make a cover song using UTAU?
To make a cover in UTAU, you assemble three things: the program itself as your "studio," a UST as the sheet music, and a voicebank as the singer. You then refine the notes and tuning, and render the result to an audio file you can mix into a track.
The basic workflow
- Get the program running. Install UTAU and open it. This is the environment where everything else happens.
- Load a UST. A UST (UTAU Sequence Text) is the project file containing the notes, BPM, aliases, pitchbends and vibrato for a song. Using an existing UST for the song you want is far faster than drawing every note by hand.
- Choose a voicebank. The voicebank supplies the actual sung samples. Load it so the UST's notes have a "singer" to perform them.
- Check the fit. The UST's aliases must match the voicebank's aliases. A Japanese voicebank reading a UST written for a different recording style or language will produce silent or wrong notes until aliases are fixed.
- Tune. Adjust pitchbends, vibrato, note timing and volume. This is where a stiff, robotic render starts sounding like a performance.
- Render. Tell the program to output the singing, then bring that audio into your music software to mix with the instrumental.
What each piece actually does
| Piece | Role | What you edit |
|---|---|---|
| UTAU | The studio/program | Project settings, rendering |
| UST | Sheet music | Notes, BPM, pitchbends, vibrato, aliases |
| Voicebank | The singer | Nothing directly — you pick and load it |
| oto.ini | Voicebank configuration | Sample start/end points and timing |
| .frq files | Pitch data | Lets UTAU flatten pitch so notes land in tune |
The voicebank's audio files are the raw sung sounds; its configuration files (oto.ini) tell UTAU where each sample begins and ends and how to time it to a note. Aliases are the names a UST uses to call up a specific sample, which is why mismatched aliases are the most common cause of a cover that loads but stays silent.
A practical example: you want to cover a fast Japanese pop song. You find a UST for it, load a Japanese CV voicebank, and immediately hear some notes cut off. The cause is usually timing in oto.ini or a pitchbend that is too sharp for the tempo, not the voicebank being "bad." Slowing the BPM temporarily while you fix the bends, then restoring it, is a reliable way to diagnose this.
If you are starting from zero, utauguides splits its material into Overview, Basic and Advanced sections, which maps well onto the order above: understand the concepts first, learn to operate the program second, then improve tuning and voicebank choice. For the program itself, UTAU Forum is a long-running community where you can find USTs, voicebanks and troubleshooting help from other users.
Decision criteria
- Pick a UST that already exists for your song unless you specifically want to transcribe it. Writing a UST from scratch teaches you a lot but is a separate skill.
- Match the voicebank to the UST's language and alias style before you spend time tuning. Fixing aliases after tuning wastes the work.
- Tune in small passes — timing first, then pitchbends, then vibrato and dynamics. Changing everything at once makes it hard to hear what helped.
What are the best practices for designing an UTAU character?
Character design in UTAU is mostly a documentation and consistency problem, not a drawing problem. The voicebank is the product; the character is its packaging, and the packaging has to stay coherent across the oto.ini, the audio samples, the readme, and any art you publish.
Start from the voice, then design the person
Decide the vocal identity before the visual one: range, tone colour, language coverage, and intended genre. A soft, breathy bank and a sharp, powerful bank imply different characters, and an audience will notice when the art contradicts what they hear. Pick a small number of traits that the voice can actually support — age impression, energy level, formality — and let everything else follow from them.
Treat metadata as part of the design
The utauguides overview explains that a voicebank pairs audio files with configuration files, including oto.ini entries that split, time and name samples, plus aliases and .frq pitch files. Design decisions land in those files whether you plan for them or not:
- Alias scheme. Choose a naming convention early (CV, VCV, CVVC, or a mix) and apply it without exceptions. Inconsistent aliases are the most common reason a UST fails to find the right sample.
- Character naming. Use one romanised name, one native-script name, and one short alias, and use them identically in the folder name, readme, and character.yaml. Mismatches confuse users and search.
- Sample coverage. Decide which phonemes and combinations the character "can say." A character presented as multilingual needs the recordings to match that claim.
Keep a character sheet that others can build on
Write down the essentials in one place: name, pronouns, age impression, height, palette, outfit variants, personality notes, and voice parameters. This is what fan artists, cover artists and collaborators actually use. Include a short "do and don't" list — for example, whether the character is depicted with a specific hairstyle or instrument. Ambiguity invites drift, and drift makes a bank feel like several unrelated characters.
Design for the medium you will actually release in
If you plan to distribute the bank, the utauguides page notes that folders, readme.txt and character.yaml matter for managing and releasing voicebanks across programs. Budget time for those files, plus a clear licence statement and usage terms. A beautiful design with unclear usage rules gets used less than a plain one with obvious permissions.
A practical next step
Sketch the character against three constraints at once: the voicebank's real range and language, the alias scheme you will record, and the licence you are willing to grant. If a design idea cannot survive all three, change the design rather than the voicebank. For terminology and workflow grounding, the beginner-oriented explanations at utauguides are a reasonable starting reference, and UTAU Forum is where community conventions and feedback on character presentation tend to surface.
How do I troubleshoot common issues when using UTAU?
Troubleshooting UTAU usually means isolating one of three layers: the program, the voicebank, or the UST. Most "it won't sing" problems trace back to a mismatch between those layers rather than a bug in the software itself.
Start with the symptom, not the software
- No sound at all, or silence on render: check that a voicebank is actually selected for the track, that the notes have lyrics or aliases filled in, and that the render output isn't muted or routed to a device that isn't connected.
- Wrong or garbled syllables: this is almost always an alias problem. The UST is asking for a sample name the voicebank doesn't have.
- Off-pitch or robotic singing: often a missing or stale frequency file, or a voicebank recorded at a different pitch than the UST expects.
- Notes that won't line up with the beat: usually a BPM mismatch between the UST and the intended tempo.
The voicebank layer
As utauguides explains, a voicebank is built from audio files plus configuration files — the oto.ini files that tell UTAU where each sample starts, ends, and sits in time. Aliases are the names that let a UST "command" a specific sample, and frequency files carry pitch information so the singer can be flattened onto the right note.
That structure explains the two most common failures:
| Symptom | Likely cause | Practical fix |
|---|---|---|
| One syllable is silent | Alias missing or misspelled in oto.ini | Check the alias list against what the UST requests |
| Everything sounds flat or wobbly | Missing or outdated .frq file | Regenerate the frequency file for the affected samples |
| Sample cuts off early | Oto.ini offset or cutoff set too tightly | Adjust the start/end points for that sample |
If a voicebank works in one UST but not another, the voicebank is probably fine and the UST is asking for something it doesn't contain.
The UST layer
A UST carries BPM, notes, aliases, pitchbends, and vibrato. When a cover sounds wrong rhythmically, compare the UST's BPM value against the original track before touching anything else. When it sounds wrong melodically, check whether the UST was written for a different voicebank type — a UST built for one recording style may not map cleanly onto a voicebank with different sample naming.
A practical order of operations
- Confirm the voicebank is selected and the project renders any sound at all.
- Play a single note in isolation. If one note works, the problem is note-specific, not global.
- Compare the UST's requested aliases against the voicebank's alias list.
- Regenerate frequency files only after aliases check out.
- Test the same voicebank in a known-good UST to decide whether the fault is the voicebank or the sequence.
For a beginner stuck on basic concepts, utauguides splits its material into overview, basic, and advanced sections, which is a reasonable way to avoid advanced fixes before the fundamentals are in place. If you want a second reference for the program itself, the official UTAU documentation and community wikis cover installation and rendering settings that a tutorial site won't always repeat.
Cloudflare
Google Workspace
User reviews (0)