﻿Amitaro's MANA Corpus Voice Recording — Punsuka  Ver.1.0  Updated: 2026-06-05

About this corpus: https://amitaro.net/voice/corpus-list/

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Contents

  1. About this corpus
  2. About this style
  3. All available styles
  4. Recording environment
  5. Update history

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

1. About this corpus

This collection contains voice recordings of all 228 sentences from the
MANA Corpus, a Japanese text corpus designed by Kana Kakonaka as a
supplement to the ITA Corpus, read aloud by Amitaro.

Format: WAV, PCM 48,000 Hz / 16-bit / Mono
        and PCM 44,100 Hz / 16-bit / Mono (both included)
Use the 44,100 Hz version when combining with dialogue voice materials.

  MANA Corpus sentence list:
    https://github.com/shirowanisan/coeiroink-corpus-manager/blob/main/mana-corpus/mana-corpus.pdf

The MANA Corpus was designed as a phonetic-balance supplement to the
ITA Corpus. It is NOT recommended for training a voice synthesis model
on MANA Corpus data alone. Please use it together with the ITA Corpus.

  ITA Corpus recordings: https://amitaro.net/voice/ita/

Recorded in a home recording studio. Each sentence recorded in multiple
takes; the most stable take was selected. Hum noise was removed by
batch processing in RX10 Advanced, followed by manual removal of lip
noise and pop noise, taking care not to suppress consonants.

Although not raw unedited speech, the audio is clean and ready for
voice synthesis applications.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

2. About this style

The "Normal" style is precisely articulated, optimized for accessibility
TTS, which can sound robotic as a character voice.

"Punsuka" (a Japanese onomatopoeia for sulking/fuming) is the sound of
an anime-style young girl working hard to be angry — noticeably irritated
but still cute, without turning hostile. Good for error messages, reactive
dialogue, or dramatic skits.

ITA Corpus and AMITARO Corpus voice recordings are also available.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

3. All available styles

"Amitaro's MANA Corpus Voice Recording" is available in the following styles.

  [All 228 MANA Corpus sentences]
  - Normal                … Precise, clearly articulated — optimized for TTS
  - Runrun                … Upbeat, bouncy delivery — like sharing good news
  - Yofukashi             … Natural, calm speech — late-night radio style
  - Punsuka (this zip)    … Animated "trying-to-be-angry" character voice
  - Sasayaki A            … Hushed whisper — heavy on breathiness
  - Sasayaki B            … Whisper with more of Amitaro's natural voice character

  Full list: https://amitaro.net/voice/corpus-list/

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

4. Recording environment

  Room                 Home recording studio
  Microphone           audio-technica AT4040 (condenser)
  Audio interface      Steinberg UR242
  Pop filter           STEDMAN PROSCREEN XL (metal)
  Headphones           audio-technica ATH-SX1a
  Reflection filter    CLASSIC PRO CAR900

The studio is a custom-built room (~2.4 tatami / ~3.8 m²) with
cellulose fiber insulation inside all walls, perforated acoustic panels
covered with thick curtains, carpet flooring, sound-deadening ceiling
panels, and an air gap between the studio floor and the ceiling below
to block noise from downstairs.

Measurement data (taken before the acoustic panels were covered with
thick curtains to reduce reverb; current reverb is lower):

  A-weighted noise level:  18.5 dB  (Rion NL-52EX)
  Reverberation time T60:  220 ms

  Measured by: Masanori Morise

Details: https://amitaro.net/about/aboutrec/

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

5. Update history

  2026-06-05  Ver.1.0  Initial release

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Amitaro's Voice Material Studio  https://amitaro.net/
