Shows the local path to a specific model. The voice_selector has the format:
${CODE}:${NAME}@${QUALITY}/${SPEAKER}
The ${SPEAKER} part is optional; as is the ${CODE} part. So if you want to select the voice named 'alba' in quality 'medium', you could simply query: alba@medium
The language code is infered. Alternatively, you can just query with the model name listed by the list command.
${CODE}-${NAME}-${QUALITY}
So for the example above, that would be en_GB-alba-medium
Path to channel (named pipe (aka. fifo)) to which piper is listening.
-j, --json Encode the text as json payload. Is on by default.
-r, --raw Encode the text directly.
-o OUTPUT, --output OUTPUT
Instead of streaming to audio channel, specifies a path to wav file where speech will be store in.
-v, --verbose Activate verbose logging.
Currently only works on linux / bsd systems, with a FIFO (aka. named pipes) setup. The basic idea is, having one pipe accepting json input (provided by this command), which is listened to by piper. After piper has processed the audio, it is either saved to file or passed on to another FIFO, which can then be read by a streaming audio player like aplay.
Example: Assuming piper is installed at /opt/wind/piper, the named pipes are located at /opt/wind/channels and whistle is available in $PATH, the aformentioned setup could look like the following:
The tail command makes sure, that the payload on speak is send to input, thereby keeping the file open after processing. Otherwise, the setup would exit after piper has finished the first payload. This way you can continually prompt.
-a, --all List voices for all available languages.
-L, --languages List available languages.
-g, --legal Show avaiable legal information.
-U, --show-url Show URL of voice on remote host.
-S, --omit-speakers Omit speakers form listing.
-p, --install-path Show path of voice (if installed).
-l LANGUAGE_CODE, --language-code LANGUAGE_CODE
Only list voices matching this language.
-i VOICE_INDEX, --voice-index VOICE_INDEX
List only specific language voice.
This command lets you investigate available voices for specific languages, or simply list all available voices. Using the –installed switch, you can filter voices that are currently installed in the local cache directory. The cache is located in the user app path, as provided by userpaths pip package. On linux this would be ${HOME}/.config/piper-whistle. You may also get the model path on the remote host using -U.
voice_index Specific language voice. (defaults to first one)
optional arguments:
-h, --help Show help message.
-v, --verbose Activate verbose logging.
-D, --dry-run Simulate download / install.
With install you can fetch available voice models and store them locally for use with piper. You may first want to search for a voice you like with list and then note the language code and index, so install knows where to look. The model file (onnx) as well as its accompanying config (json) file, will be stored in the local user data path as provide by userpaths. On linux this would be ${HOME}/.config/piper-whistle.