Piper-phonemize for Windows?

[JohnyB7]

I'd love to try out Piper, but I'm not sure how to install it on Windows 10 as the pip install requires piper-phonemize and I'm not sure how to get it running on Win10.

[StoryHack]

If you have a relatively modern machine you can use windows subsystem for linux (wsl) to install and run piper. It basically makes a virtual linux machine for you. You do need to learn a bit about using wsl, though, so it's not so straightforward and just having a windows distribution.

[JohnyB7]

Thanks you for the explanation, but sadly this is far too complicated for me and I don't have enough free time to learn how to do it.
I will simply use Tortoise TTS instead of Piper.
I don't know why the developer decided to make the Piper project dependent on something that isn't compatible with 95% of the OS market and then abandon the project.
I hope someone forks the project and makes it more accessible to the vast majority of potential users.

[synesthesiam]

Work is being done on a Windows port. I'm the only dev, and I work in Linux so that has been the target platform.

Windows is a pain to support because I can't just pull a Docker build image and create an automated workflow. I have to use a VM or GitHub actions since cross-compilation hasn't worked either.

[Feanix-Fyre]

I tried tackling the GitHub Actions compilation from where you got stuck on piper-phonemizer. You can check my fork for how far I got. One issue was that windows uses slashes in the opposite direction from Linux so I swapped those out. Another issue was that it couldn't find something so I told it where it was directly. And then I got stuck cause I didn't actually know what I was doing.

[synesthesiam]

A Windows pre-release is now available for testing: https://github.com/rhasspy/piper/releases/tag/2023.9.9-1

[Feanix-Fyre]

I had an issue with json as an input, but it does perform its MVP (double entendre) for what it claims: lightweight, faster than real time, easy to use, easy to train, high quality TTS model. I just call it with a batch or python script if I want to do a bunch of inference. I want to recompare to see if there is any difference in how it is breaking up long sentences and how the colab/linux version would but not necessarily within the scope of the project or your intentions as a developer. My evaluation is that its intention is to be for the edge so this windows port is just a blessing and I'll take it as it is. With how lightweight and fast it is, it's not a problem for it to be called over and over. Not necessarily the most efficient but I was working with Tortoise before this so....... :) I'm a happy camper. Thank you, kindly, for the hard work.

[Kudos6505]

Did you get the json or audio streaming to work? I worked around the json input by using a txt file and "type" command. I still don't know how to stream audio, as aplay isn't on windows.

[muhammad-meraj]

I am trying to use piper in the project which runs on arm-wince device. If there is any compiled version or binary for arm-wince device?
Or any guide to compile it for arm-wince device. Thanks

[csukuangfj]

https://github.com/k2-fsa/sherpa-onnx
uses piper-phonemize and it supports Linux, windows, macOS, iOS, Android.