Skip to main content

Synthese vocale par concatenation de diphones WORLD — francais

Project description

lectura-tts-diphone

Synthese vocale francaise par concatenation de diphones dans le domaine WORLD.

Installation

# Sans dependances (import seul)
pip install lectura-tts-diphone

# Inference locale (pyworld + numpy + scipy)
pip install "lectura-tts-diphone[local]"

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

Utilisation

Depuis du texte (necessite lectura-phonemiseur)

from lectura_tts_diphone import synthetiser

audio = synthetiser("Bonjour le monde")
# audio: numpy array float32, 44100 Hz

Depuis des phonemes IPA

from lectura_tts_diphone import creer_engine

engine = creer_engine()
audio = engine.synthesize_groups([
    {"phones": ["b", "ɔ̃", "ʒ", "u", "ʁ"], "boundary": "none"},
    {"phones": ["l", "ə", "m", "ɔ̃", "d"], "boundary": "period"},
])

Controles prosodiques

Parametre Defaut Description
duration_scale 1.0 Vitesse globale (>1 = plus lent)
pause_scale 1.0 Duree des pauses inter-groupes
macro_expressivity 2.0 Gestes prosodiques (0=neutre, 4=exagere)
micro_expressivity 5.0 Micro-variations (0=robot, 10=tres expressif)
spectral_contrast 1.5 Contraste spectral (1.0=off, 2.0=fort)
prosody_style "auto" "declaratif", "question", "exclamation", "suspensif", "neutre"
seed None Graine pour micro-prosodie reproductible

Modes de synthese

  • FLUIDE : lecture naturelle, enchainement continu
  • MOT_A_MOT : lecture mot par mot avec pauses
  • SYLLABES : lecture syllabe par syllabe

Architecture

Texte → [G2P] → Phonemes IPA → Diphone chain
                                      ↓
                              WORLD params (F0 + SP + AP)
                                      ↓
                              Stretch + Concat (overlap)
                                      ↓
                              Prosodie (F0 contour + durees)
                                      ↓
                              GV compensation (contraste spectral)
                                      ↓
                              pw.synthesize → Audio 44100 Hz

Les diphones sont des parametres WORLD (F0 + spectral envelope + aperiodicity) extraits du corpus SIWIS et moyennes par type de transition phonetique.

Emplacements des modeles

Recherche dans l'ordre :

  1. Parametre models_dir explicite
  2. $LECTURA_MODELS_DIR/tts_diphone/
  3. ~/.lectura/models/tts_diphone/
  4. Modeles embarques dans le package

Fichier requis : diphones.dpk.gz (ou .dpk.gz.enc chiffre) Fichier optionnel : diphone_statistics.pkl

Licence

Double licence : AGPL-3.0 (code) + Licence Commerciale (modeles).

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

lectura_diphone-2.0.0.tar.gz (87.5 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

lectura_diphone-2.0.0-py3-none-any.whl (81.1 kB view details)

Uploaded Python 3

File details

Details for the file lectura_diphone-2.0.0.tar.gz.

File metadata

  • Download URL: lectura_diphone-2.0.0.tar.gz
  • Upload date:
  • Size: 87.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.11.9

File hashes

Hashes for lectura_diphone-2.0.0.tar.gz
Algorithm Hash digest
SHA256 4c558b27cb5eb1c6324bda535192627f3ec0706c866b2828851e5134ae76adf8
MD5 0201255edc9e1fa8b8992be22ede1045
BLAKE2b-256 20f327504735b99d6e1c246952e2474d60e8871533f976d51fcd2e42839c7251

See more details on using hashes here.

File details

Details for the file lectura_diphone-2.0.0-py3-none-any.whl.

File metadata

File hashes

Hashes for lectura_diphone-2.0.0-py3-none-any.whl
Algorithm Hash digest
SHA256 a365d7880c707bf23bbd3dbdff81c187ef072348d69cd8a88cf3278dde77de90
MD5 e907c61a968be393fa0aea9c8bc7ddd9
BLAKE2b-256 d67616a93f3eeba5645ba57145a9d486721007807d59792d7d2b8b65327a8b98

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page