lectura-tts-monospeaker 1.3.3

Synthese vocale neuronale monospeaker francais — FastPitch-Lite + HiFi-GAN (ONNX)

lectura-tts-monospeaker

Synthese vocale neuronale monospeaker pour le francais.

Installation

# Version minimale (API distante, zero deps)
pip install lectura-tts-monospeaker

# Version locale (inference ONNX)
pip install lectura-tts-monospeaker[onnx]

# Avec G2P integre (texte → audio)
pip install lectura-tts-monospeaker[all]

Utilisation

Depuis du texte (necessite lectura-g2p)

from lectura_tts_monospeaker import synthetiser

audio = synthetiser("Bonjour le monde")
# audio: numpy array float32, 22050 Hz

Depuis des phonemes IPA

from lectura_tts_monospeaker import creer_engine

engine = creer_engine(mode="local")
result = engine.synthesize_phonemes(
    "bɔ̃ʒuʁ",
    phrase_type=0,
    pitch_range=1.3,
)
# result.samples: numpy float32
# result.sample_rate: 22050
# result.phoneme_timings: list[PhonemeTiming]

Via l'API distante

from lectura_tts_monospeaker import creer_engine

engine = creer_engine(mode="api", api_url="https://api.lec-tu-ra.com")
result = engine.synthesize("Bonjour")

Controles prosodiques

Parametre	Defaut	Description
duration_scale	1.0	Vitesse globale
pitch_shift	0.0	Decalage F0 (demi-tons)
pitch_range	1.3	Variation F0 (1.0 = neutre)
energy_scale	1.0	Intensite
pause_scale	1.0	Duree des pauses
phrase_type	0	0=decl, 1=inter, 2=excl, 3=susp

Architecture

• FastPitch-Lite : phonemes → mel-spectrogramme (~5M params)
• HiFi-GAN V1 : mel → audio 22050 Hz (~3.5M params)
• Runtime : ONNX (pas de dependance PyTorch)

Licence

Licence proprietaire Lectura. Voir LICENCE.txt.