0.89.0
This release has one of the first AI features! We use sherpa-onnx and and automatically download models.
For English, vits-piper-en_US-libritts_r-medium is probably a good choice, or use kokoro-multi-lang-v1_0 if you don't need high performance.
You can add a new generated .mp3 by using the module Quick Actions.
To use it in Chandler, create a config entries resource with the key core_plugin_tts/default_model.
Then you can use the "speak" command block. In the future, there will probably be more features using TTS.
Piper and Kokoro are supported, but Kokoro is much slower than realtime on
edge devices, so it may not be usable. Generated speech is cached compressed in RAM, so you
might be able to use it if you have a few repeated phrases.
- ✨ New system tag points and default alarms for hdmi port connection(/system/display_ports/hdmi_1.connected)
- ✨ Stateful blocks like on_change in the visual logic editor.
- ✨ Speak command that useds the new globally configured TTS model.
- 🐛 Fix web media not rendering without login
- 🐛 Fix mixer sort issue and use fixed width num param fields
- ✨ Ability to recover from corrupted chandler state database
- 🐛 Useless nuisnce vignette warning logs
- ✨ Use plain exception formatting when not running in a TTY
- ✨ Audio features work correctly when installed via UV