響 Hiro UST

Procedural UTAU .ust generator for Japanese lyrics with mora-accurate parsing, motif memory, voice leading, and musical scales. Built by Ilya Minin (Eli) - interactive installations & generative audio specialist.

🎛️ Purpose

Converts Romaji/Hiragana lyrics → UST files with procedural melodies that respect:

• Japanese phonetics (mora boundaries, gemination っ, yoon)
• Musical structure (phrase endings, motif repetition)
• Voice ranges (Soprano→Bass)

🎵 Usage

[Verse]
きゃっきゃ うれし いたい さぶり
ゆびさき きりさけ あかい つゆ

[Chorus]  
いたみ いたみ きもちいい

→ Single-click UST generation with melody, timing, dynamics.

✨ Core Mechanics

1. Mora Trie Parser

Romaji → Hiragana → Phonemes (O(1) lookup)
っか → ['っ','ka'], きゃ → ['kya']

• Small tsu gemination (っ)
• Full yoon support (kya, sha, nya)
• Vowel/consonant timing differentiation

2. Procedural Melody Engine

Voice Leading: Tight(±1) → Wild(±5) semitones
Motif Memory: Learns/reuses 4-note patterns (40% prob)
Phrase Structure: Resets to tonic/dominant every 6-12 notes
Chord Awareness: I-IV-V progression cycle

3. 20+ Scales

- Major/Minor Pentatonic (Japanese traditional)
- Diatonic majors/minors (C→A)
- Whole Tone, Octatonic (generative/experimental)
- Chromatic, Blues, Hexatonic

🚀 Get Started

EXE (Recommended)

# Download from Releases
HiroUST_v4.1.exe → Double-click → Generate UST

Python

pip install tkinter  # Usually pre-installed
python hiro_ust.py

🎚️ Controls

Section	Params	Effect
⏱️ Timing	Tempo, Base Length, Line/Section Pauses	Song structure
🎤 Voice	Soprano(67)-Bass(48), Scale	Pitch range + palette
🎵 Behavior	Motif Memory, Lyrical Mode, Microtones, Chords	Melodic character
⚙️ UST	PreUtterance, Overlap, Intensity, Envelope	Rendering quality

🎼 Modes

• 🎼 Motif Memory - Remembers + varies 4-note patterns
• 🎭 Lyrical - Vowels ↑ high, Consonants ↓ low
• ♯ Microtones - Quarter-tone vowel bends
• 🎸 Chords - I(0-2), IV(3-4), V(5-7) beat cycle
• 📏 Intone - Tight(1)→Wild(5) leap control

💾 Presets

Save/load complete configurations:

Pop_Idol.json, Horror_Ambient.json, Experimental.json

🧮 Technical

Parser: Trie-based O(n) mora lookup
Melody: Semi-procedual, semi-random
State: Persistent phrase/motif memory
Output: UTF-8-sig UST (UTAU v1.2 compatible)

🎨 By Ilya Minin (Eli)

Background: Interactive installations, generative audio (MAX/MSP) → UTAU, SynthV and Japan)))

Eli_lab - Moscow-based contemporary artist & creative technologist

License: MIT - Free for all uses