ondrazarici/kasandra-local-voice-cz-piper — explained in plain English
Analysis updated 2026-05-18
Add a natural-sounding Czech voice to a Home Assistant smart home setup running Piper.
Run Czech text-to-speech entirely offline without sending any audio to the cloud.
Replace a default or robotic-sounding Piper voice with a purpose-built Czech female voice.
| ondrazarici/kasandra-local-voice-cz-piper | 0marildo/imago | 0xvasic/lobby-radio-player | |
|---|---|---|---|
| Stars | 3 | 3 | 3 |
| Language | — | Python | Visual Basic .NET |
| Setup difficulty | easy | easy | easy |
| Complexity | 1/5 | 2/5 | 1/5 |
| Audience | general | general | general |
Figures from each repo's GitHub metadata at analysis time.
Requires purchasing the voice package separately before it can be installed, installation itself is just copying files into Piper's voices folder.
Kasandra is a Czech-language female voice model built for Piper, a text-to-speech engine, aimed at bringing natural-sounding Czech speech to smart home setups such as Home Assistant. It runs entirely offline on your own hardware, so no audio or data is ever sent to a cloud server. The README explains the motivation plainly: most voice assistants either do not support Czech at all, or require an internet connection and send your voice data to large companies' servers. Kasandra is offered as an alternative that keeps everything local, protects privacy since nothing leaves your home network, responds quickly because it is tuned to work well with Piper, and aims to sound natural and easy to understand in Czech. This is not free open source software you clone and build. Building and training a custom voice model takes real time and resources, so the author sells the complete package, including its license terms, through a linked Ko-fi page rather than distributing it in this repository. The repository itself mainly serves as the description and installation guide. Technically, Kasandra is distributed as an ONNX model file compatible with Piper, intended especially for use with Home Assistant through its Piper or Wyoming protocol integration. Installation involves downloading the purchased package, placing the voice files into the voices folder used by your Piper or Home Assistant setup, and then selecting Kasandra as the active voice in settings. For anyone building a smart home around privacy, the appeal here is straightforward: Czech is not always well supported by mainstream voice assistants, and the options that do support it often mean sending recordings of your household to a remote server. Kasandra sidesteps that entirely by running as a local model file that Piper loads and speaks from directly, with no ongoing internet connection needed once it is installed. The README frames this project as the first release from its author rather than a large commercial product, and it is priced and distributed accordingly, as a single downloadable package rather than an open catalog of voices. The author invites questions or improvement suggestions through GitHub issues on this repository, and treats the project as something that may keep evolving based on feedback from people actually using it in their homes.
A paid, offline Czech female voice model for the Piper text-to-speech engine, built for use with Home Assistant.
The voice model is not free or open source, it must be purchased separately through the author's Ko-fi page, with its own license terms included in that purchase.
Setup difficulty is rated easy, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
double-check against the repo, no cap.