DOCS/CONTENT & AUDIO CRAFT/CMU Arpabet Phonemes

CMU Arpabet Phonemes

Accurate phonetic overrides for anime lore, fantasy names, and uncommon terms.

CMU Arpabet phonetic overrides guarantee that neural voice synthesis pronounces names, foreign words, and anime lore accurately on the very first take.

Why Phonemes Matter

Neural TTS engines frequently mispronounce unfamiliar names (e.g. "Frieren", "Zoltraak", "Arkhipov", "Carbajal"). Stage 02 enforces phonetic overrides: shorts verify-script fails if any entity in final.md lacks an Arpabet mapping in config.yaml.

CMU Arpabet Cheat Sheet

Arpabet codes represent sounds in 2-letter uppercase symbols. Vowels end in stress numbers (0 = unstressed, 1 = primary stress, 2 = secondary stress):

VOWELS & DIPHTHONGS

AA1 : balm

AE1 : bat

AH0 : about / butt (AH1)

EH1 : bet

EY1 : bait

IH1 : bit

IY1 : beet

OW1 : boat

UW1 : boot

CONSONANTS

CH : chin

DH : this

HH : hat

JH : joy

NG : sing

SH : she

TH : thin

ZH : pleasure

Real-World Examples

series/<series>/config.yaml

1phonemes:

2 # Anime / Fantasy Entities

3 Frieren: F R IY1 R AH0 N

4 Himmel: HH IH1 M AH0 L

5 Zoltraak: Z OW0 L T R AA1 K

6 Flamme: F L AA1 M

7 Amon: AA1 M AA0 N

9 # Real-World Historical Entities

10 Hicks: HH IH1 K S

11 Carbajal: K AA0 R B AH0 HH AA1 L

12 Arkhipov: AA0 R K IY1 P AO0 V

13 Yamaguchi: Y AA2 M AH0 G UW1 CH IY0

Was this "CMU Arpabet Phonemes" page helpful?

High-Retention Scripting

The 100-130 word formula: hooks, emotional pacing, banned AI tell words, and loop closes.

Image Scraping & Vision

Multi-source retrieval, OCR watermark rejection, framing tags, and library grammar.