Repository navigation
This release updates the underlying ai-coustics SDK to 0.26.0.
Breaking Changes
New model file version
This release requires model file version 8.
Re-download your models so the SDK does not reject them with a "model version is not supported" error.
See the compatibility matrix.
New Features
-
Quail Voice Focus 2.2 has been replaced by version 2.3, with model key
quail-vf-2.3-l-16khzandquail-vf-2.3-s-16khz.
Quail Voice Focus 2.2 (quail-vf-2.2-l-16khz/quail-vf-2.2-s-16khz) is not loadable by this SDK version any more. -
VAD Voice Focus 2.0 has been replaced by version 2.1, with model key
vad-vf-2.1-s-16khz.
VAD Voice Focus 2.0 (vad-vf-2.0-s-16khz) is not loadable by this SDK version any more. -
Tyto 1.1 has been replaced by version 1.2, with key
tyto-1.2-l-16khz.
Tyto 1.1 (tyto-1.1-l-16khz) is not loadable by this SDK version any more.
Improvements
- The SDK now supports loading models with half-precision parameters (
f16),
improving inference speed on memory-bound models without sacrificing model performance.