i'm trying to train my own model, but it's failing installing piper-phonemize. it seems like that repo is more or less dead as well. have we moved past piper and are over at using something else now?
#is piper dead? last release on the repo is nearly two years ago?
subtle onyx
i'm following this guide: https://github.com/rhasspy/piper/blob/master/TRAINING.md (which also is over two years old) and getting an error when installing the requirements.
Obtaining file:///root/src/piper/src/python
Preparing metadata (setup.py) ... done
Collecting cython<1,>=0.29.0 (from piper_train==1.0.0)
Using cached Cython-0.29.37-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.manylinux_2_28_x86_64.whl.metadata (3.1 kB)
INFO: pip is looking at multiple versions of piper-train to determine which version is compatible with other requirements. This could take a while.
ERROR: Could not find a version that satisfies the requirement piper-phonemize~=1.1.0 (from piper-train) (from versions: none)
ERROR: No matching distribution found for piper-phonemize~=1.1.0
Too new version of python?
paper slate
Why not using the pre build docker version ?
sinful hamlet
The root cause is unfixed dependency versions. The team has no time for updates, and initially, Piper was essentially a one-person project. Since there are a large number of English-language tts projects, this does not seem to be a big problem. But for many languages, this is the only available local option. However, there are ready-made enthusiast projects for voice training https://github.com/domesticatedviking/TextyMcSpeechy. Adding a new voice to HA is yet another challenge.
subtle onyx
is textymcspeechy supposed to simplify the process of running piper?
sinful hamlet
is textymcspeechy supposed to simplify the process of running piper?
These are training tools, a Docker container with a working set of code. You can also run Piper in Docker, as the HA add-on does.
sinful hamlet
You can also pay attention to a new project. Especially for training. https://github.com/OHF-Voice/piper1-gpl/blob/main/TRAINING.md
subtle onyx
i tried training the model earlier today, but it came out terrible. any suggestions? i'm speaking in a dialect, and we pronounciate some words quite different from the written version