g2p2 0.1.1

Zero-dependency grapheme-to-phoneme for the 100 Whisper languages (Rust core)

g2p2

Zero-dependency grapheme-to-phoneme for the 100 Whisper languages, with a Rust core exposed to Python.

pip install g2p2

from g2p2 import Model

m = Model.load("fr.g2p")
print(m.phonemize("bonjour"))            # bɔ̃ʒuʁ
print(m.phonemize_many(["chat", "eau"])) # ['ʃa', 'o']

Model blobs (*.g2p) are built from WikiPron / epitran / LLM data — see the main repo. Download a language's .g2p from the GitHub releases or build with cargo run -p xtask -- build-all.

API

• Model.load(path) -> Model — load a .g2p file
• Model.from_bytes(bytes) -> Model — load from raw bytes
• Model.phonemize(word) -> str — IPA for one word
• Model.phonemize_many(words) -> list[str] — IPA for many

License: MIT OR Apache-2.0.