|
--- |
|
language: |
|
- be |
|
--- |
|
|
|
# Bel-TTS |
|
|
|
## Install |
|
Install TTS with pip: |
|
``` |
|
pip install TTS |
|
``` |
|
You would also need to download files from this repo. |
|
|
|
## Running synthesizer |
|
### Preprocessing text |
|
This TTS model uses phonemized text as an input. To get phonemized version of the data, |
|
you need to send the text you want to voice to the phonemizer api: |
|
``` |
|
curl --location 'fonemizer.nikuchin.fun/processText' \ |
|
--header 'Content-Type: text/plain' \ |
|
--data 'Гепарды жывуць у адкрытых і прасторных месцах, дзе ёсць шмат здабычы.' |
|
``` |
|
You'll get a response with the phonemized data, that will look something like this: |
|
``` |
|
Ґеҁпарды җыҁѵуЦ ҁу аҁткрытыХ ҁйі праҁсторных ҁӎесцах, ҁЗе ҁйоСЦ ҁшмад Җдаҁбычы. |
|
``` |
|
|
|
### Synthesizing |
|
|
|
Example for the synthesizing command: |
|
``` |
|
tts --text "Ґеҁпарды җыҁѵуЦ ҁу аҁткрытыХ ҁйі праҁсторных ҁӎесцах, ҁЗе ҁйоСЦ ҁшмад Җдаҁбычы." \ |
|
--config_path ${PATH_TO_FILE}/config.json \ |
|
--model_path ${PATH_TO_FILE}/model.pth \ |
|
--out_path ${PATH_TO_FILE}/output.wav \ |
|
--vocoder_path ${PATH_TO_FILE}/vocoder.pth \ |
|
--vocoder_config_path ${PATH_TO_FILE}/vocoder_config.json |
|
``` |
|
|
|
(change ${PATH_TO_FILE} to your directory) |
|
|
|
|
|
### Fonemizer for local deployment |
|
|
|
https://hub.docker.com/repository/docker/jhlfrfufyfn/be-fonemizer-for-tts/general |
|
|
|
Pulled like this: |
|
|
|
docker pull ... |
|
|
|
docker run -d -p 8080:8080 ... |
|
|
|
And after that use endpoint localhost:8080/processText |
