KENZY: Text-to-Speech (TTS)
Parameters
Parameter | Type | Default | Description |
---|---|---|---|
group | str | None | Device membership group |
location | str | None | Location, e.g. "Living Room" |
model.type | str | speecht5 | Options are: festival, speecht5 |
model.target | str | gpu | SpeechT5 options: gpu, cpu |
speaker | str | slt | SpeechT5 options: slt, clb, bdl, ksp, rms, jmk |
external_player | str | None | External player command |
cache.folder | str | ~/.kenzy/cache/speech | Folder for caching spoken phrases |
offline | bool | false | Will disable downloading the models |
The model type of speecht5
uses the microsoft/speecht5_tts model from Huggingface.co. The festival option calls the external festival program.
Note: You should consider only setting offline
after you have executed the program at least once so that it fully downloads all model files. Once they are downloaded you can switch the offline mode on so that it does not try to re-download the models (which enables the program to then run without an Internet connection).
The external_player
option allows you to specify a program like paplay
that can play wave files rather than using the built-in pythonic player.
Example YAML file
See Service Settings for options in the service group.
type: kenzy.tts
device:
location: Living Room
group: Primary
model.type: speecht5
model.target: cpu
speaker: slt
cache.folder: ~/.kenzy/cache/speech
service:
host: 0.0.0.0
port: 9702
Help & Support
Help and additional details is available at https://kenzy.ai