xphoneBR is a library for grapheme to phoneme conversion for Brazilian Portuguese based on Transformer models.
It is intended to be used in text-to-speech production systems with high accuracy and efficiency.
You can choose between a forward Transformer model (trained with CTC) and its autoregressive
counterpart. The former is faster and more stable while the latter is slightly more accurate.
- G2P
- Normalization
- Bert Tokenizer
The library has been tested on Python 3.7 to 3.11.
To install the latest version of xphoneBR, you can use pip:
pip install xphonebr -UOr if you want to install from source, you can use:
pip install git+https://github.com/traderpedroso/xphoneBR.gitfrom xphonebr import Phonemizer
phones = Phonemizer(normalizer=True)
phones.phonemise("Dra. Ana tem 45% da empresa & cia e iniciou as 8:45 de quinta feira do ano 2024 etc. ")
output: 'dowˈtoɾə. ˈãnə ˈtẽ kwaˈɾẽtə ˈi ˈsĩkʊ ˈpox ˈsẽtʊ ˈda ẽˈpɾɛzə ˈi kõpãˈiə ˈi inisiˈow ˈas ˈoytʊ ˈɔɾəs ˈi kwaˈɾẽtə ˈi ˈsĩkʊ miˈnutʊs ˈdʒi ˈkĩtə ˈfeyɾə ˈdʊ ˈãnʊ ˈdoys ˈmiw ˈi ˈvĩtʃɪ ˈi ˈkwatɾʊ ˈit seˈteɾə.'from xphonebr import Phonemizer
phones = Phonemizer(autoreg=True, normalizer=True)
phones.phonemise("Dra. Ana tem 45% da empresa & cia e iniciou as 8:45 de quinta feira do ano 2024 etc. ")
output: 'dowˈtoɾə. ˈãnə ˈtẽ kwaˈɾẽtə ˈi ˈsĩkʊ ˈpox ˈsẽtʊ ˈda ẽˈpɾɛzə ˈi kõpãˈiə ˈi inisiˈow ˈas ˈoytʊ ˈɔɾəs ˈi kwaˈɾẽtə ˈi ˈsĩkʊ miˈnutʊs ˈdʒi ˈkĩtə ˈfeyɾə ˈdʊ ˈãnʊ ˈdoys ˈmiw ˈi ˈvĩtʃɪ ˈi ˈkwatɾʊ ˈit seˈteɾə.'from xphonebr import normalizer
normalizer("Dra. Ana tem 45% da empresa & cia e iniciou as 8:45 de quinta feira do ano 2024 etc. ")
output: 'doutora. Ana tem quarenta e cinco por cento da empresa e companhia e iniciou as oito horas e quarenta e cinco minutos de quinta feira do ano dois mil e vinte e quatro et cetera.'We welcome any contribution to xphoneBR. Here are some ways to contribute:
- Report issues or suggest improvements by opening an issue.
- Contribute with code to fix issues or add features via a Pull Request.
Before submitting a pull request, please make sure your codes are well formatted and tested.
I would like to express my gratitude to @as-ideas for creating the initial project. Their work has been an invaluable starting point for my modifications and improvements for Brazilian portuguese.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent