VAP // Voice Lab 01

Recording Booth

Loading the booth…
0 / 220 takes Stored on this device
TAKE 001 / 220 00000.wav Not recorded
Vowel warm-up

Loading prompt…

Provided reference ODiE
00:00.000
Mic check Press Record or R to begin
CLIP
Peak— dB Pitch— Length—
Everything stays in this browser until you download or erase it.
Private by design. Microphone audio never leaves this browser. WAV · mono · 44.1 kHz · 16-bit PCM
VAP Voice Lab

Your browser is the booth.

Record the supplied 220-line ARPAsing list, compare each line with the provided reference performance, and keep every take locally until you package the voicebank.

01Use headphones

Prevent the guide and reference from leaking into your microphone.

02Allow the mic

Recording begins only after you press Record. Nothing is uploaded.

03Download a backup

Browser storage can be cleared. Package your work often.

Content note: a few supplied pronunciation glosses include adult language.
Recording guide

OREMO muscle memory, rebuilt for the web.

Fast controls

R / 0
Start or stop recording
Space
Play your current take
↓ / F6 / 2
Next prompt
↑ / F7
Previous prompt
O
Play the tuning tone
M
Toggle the metronome
Ctrl / ⌘ + G
Search the reclist

How to sing the list

Use the readable gloss as an approximation and the ARPAbet line for precision. Sing each three-syllable rhyming string continuously, like a short VCV phrase. In the phoneme line, q—shown as an apostrophe in the gloss—means a brief glottal stop.

The first ten entries are vowel strings. Listen first, choose a comfortable pitch, keep your distance from the microphone consistent, and retake clipped or noisy lines.

Three recording modes

Manual records until you stop. Guide once follows the original kit’s 1.25–4.75 second recording window. Guide loop saves, advances, and cues the next missing line automatically.

What the monitor shows

Wave displays the take shape; Spectrum shows frequency energy; Pitch estimates the sung fundamental. They are coaching views, not OTO parameters, and they never leave your device.

Sample Lab edits the WAVs themselves

Sample Lab can bake optional gender/formant, tension, openness, breathiness, brightness, and unvoiced-consonant changes into one take or many stored takes. Gender is a formant/timbre control: it does not infer identity and does not move the sung pitch. The processor preserves timing and filenames, protects a −1 dBFS ceiling by default, and keeps one exact-byte undo batch until you record, import, retake, delete, or apply again. It never creates or changes oto.ini.

After download: generate the OTO with moresampler

  1. Unzip the Recording Booth package. Confirm the WAV files, index.csv, and recording-list files are together in the voicebank folder.
  2. On Windows, drag the whole voicebank folder onto moresampler.exe.
  3. When moresampler asks for the voicebank type, enter 3 for ARPAsing and press Enter.
  4. When asked whether to number duplicate diphones, answer y or yes.
  5. On macOS or Linux with Wine, run wine moresampler.exe "/path/to/voicebank" and answer the same prompts.
  6. Back up and review the generated oto.ini in SetParam, UTAU, or OpenUtau. Automatic timing is a starting point, not a finished OTO.

moresampler is a separate third-party tool and is not bundled with Recording Booth. See the community-maintained ARPAsing voicebank tutorial for the maintained workflow.
Voicebank package

Pack the session.

Your recorded WAVs will be bundled with the matching reclist, comments, index, metadata, original guide MIDIs, and setup directions.

No OTO is generated here. The download includes exact moresampler steps. A partial bank can still be packaged, but a complete 220-take bank is recommended.
Sample Lab · local DSP

Shape the WAVs themselves.

Audition voice shaping, then render it directly into the locally stored samples used by playback and package export.

Reversible once: Apply replaces stored WAV audio—not playback flags or OTO data. One exact-byte undo batch is kept locally. A later record, import, retake, deletion, or edit clears that undo.

01 Choose the batch

Record or import at least one take to use Sample Lab.

02 Starting tone
Presets only move the controls. Nothing is stored until Apply.
03 Voice shaping
Local data

Erase every take?

This permanently removes this Recording Booth session from this browser. Download a package first if you need a backup.
People & provenance

Built on a community of voice makers.

Recording Booth is an independent, clean-room web implementation. It is not affiliated with UTAU, ARPAsing, OREMO, or moresampler, and it does not copy or bundle those programs or their source code.

01

UTAU platform

Ameya/Ayame (飴屋／菖蒲) created UTAU and UTAU-Synth. Recording Booth prepares a classic UTAU-style voicebank folder but does not include or modify UTAU.

Official UTAU site ↗
02

ARPAsing method

Kanru Hua designed ARPAsing, a systematic method for English singing synthesis in UTAU. ARPAsing is a method rather than one immutable reclist; its 0.1.0 proposal established the canonical 220-sample structure behind this kit.

ARPAsing method & attribution ↗
03

Supplied reclist edition & kit adaptation

The project owner attributes this supplied indexed edition, OREMO comments, and recording-kit preparation to Adlez27 (also DystoP/KLAD). Their site requests the Adlez27 name for non-music work. The public ARPAsing directory credits the underlying canonical 0.1.0 220-sample list to Kanru Hua, so both roles are preserved here.

Adlez27 / DystoP’s site ↗ · ARPAsing reclist lineage ↗
04

Reference performance

ODiE. All 220 references are encoded from the ODiE V1 Alpha ARPAsing voicebank by Seiun—the same ODiE bank used by BloomBeat.

05

OREMO inspiration

nwp8861 created OREMO, the UTAU voicebank recorder whose reclist navigation, keyboard workflow, analysis views, tuning tone, metronome, and guide-recording modes informed this browser design. No OREMO code is shipped here.

OREMO creator listing ↗
06

moresampler & automatic OTO workflow

Kanru Hua created moresampler. Recording Booth intentionally leaves OTO generation to that separate tool and follows the community-maintained ARPAsing workflow for mode 3. A generated OTO is only a starting point and should be reviewed.

moresampler creator listing ↗ · Voicebank workflow ↗
07

ARPAsing documentation community

The linked ARPAsing resource site says it was primarily written by KLAD, with design by Tomato Hentai and responsive design by VocAddict. It is a community resource and explicitly not an official Kanru Hua site.

Site provenance ↗
08

ARPAbet notation

The displayed ASCII phoneme labels use ARPAbet-style notation, the speech-research alphabet from which ARPAsing takes its name. Carnegie Mellon’s CMU Pronouncing Dictionary is a longstanding reference for this symbol set; no CMUdict word data is bundled here.

CMUdict source & acknowledgement ↗
09

Original Booth Pulse guides

New 120 BPM MIDI compositions and clean synthesized WAV renditions created for Local Robot’s Recording Booth. They preserve the supplied kit’s recording cadence without redistributing its legacy guide music.

10

Browser platform & implementation

Designed and implemented by Local Robot. Microphone capture uses the open Media Capture and Web Audio APIs; local binary storage uses IndexedDB; package creation is implemented in-app without a third-party ZIP library. Sample Lab’s worker-based spectral-envelope DSP is likewise a dependency-free, clean-room implementation; no third-party audio-processing code is bundled.

Browser microphone API ↗
Rights note

The reclist edition and reference recordings were supplied for this project. Their creators retain their respective rights; no additional license or endorsement is claimed by Recording Booth. Your downloaded package contains your own voice recordings plus the supplied text mapping needed for the requested workflow. Confirm redistribution permission before publishing that mapping or the provided references elsewhere.
More Games by Local Robot