Arch Linux User Repository

As @evorster reported, a split package for such a big collection is not viable in the AUR.

Split packages in the repos (e.g. tesseract-data from which I took inspiration) work fine because the end user can download and install them individually, while split packages from the AUR must build and package everything, and only then the user can install only what he needs.

Therefore I am going to change the pkgname=() array and prepare_*() function from piper-voices, in order to introduce individual packages for each locale.

The issue for this and other localised piper-voices packages. The proper way to fix this is to change the following line from:

 git lfs install 

to:

 git lfs install --local 

The default behaviour of git lfs install is to setup the LFS filters inside the global .gitconfig file. Adding --local makes it modify (or create) a .gitignore file inside the current repository, which in this case is the local copy of the piper-voices repo from huggingface.co, so it doesn't conflict with any constraints enforced by makepkg.

As @evorster reported, a split package for such a big collection is not viable in the AUR.

Split packages in the repos (e.g. tesseract-data from which I took inspiration) work fine because the end user can download and install them individually, while split packages from the AUR must build and package everything, and only then the user can install only what he needs.

Therefore I am going to change the pkgname=() array and prepare_*() function from piper-voices, in order to introduce individual packages for each locale.

Seems like the files are not being pulled correctly as well with the current PKGBUILD. Does it work for you @evorster ?

>cat /usr/share/piper-voices/en/en_GB/alan/medium/en_GB-alan-medium.onnx                                                       
version https://git-lfs.github.com/spec/v1
oid sha256:0a309668932205e762801f1efc2736cd4b0120329622adf62be09e56339d3330
size 63201294

Hi there!

I can see that these packages are still in a bit of flux, but I have to report that currently it is not possible to install piper-voices-en-gb and piper-voices-en-us concurrently.

Each of those also seem to want to clone all of the piper-voices git, which kind of defeats the purpose of splitting apart all these packages....

In the piper-voices.install in line 38, there is a stray ` (backtick) that causes a syntax error.

I think there's an issue with your PKGBUILD. It ends in an error: error: command failed to execute correctly

I also tried to run the speak example command afterwords and it complains with:

terminate called after throwing an instance of 'std::runtime_error'
  what():  Model file doesn't exist
ALSA lib pcm_dmix.c:1000:(snd_pcm_dmix_open) unable to open slave
aplay: main:834: audio open error: No such file or directory
zsh: done                           echo "Hello world" |
zsh: IOT instruction (core dumped)  piper-tts --model  --output-raw |
zsh: exit 1                         aplay -r 22050 -f S16_LE -t raw -

The audio file generation with:

terminate called after throwing an instance of 'std::runtime_error'
  what():  Model file doesn't exist