Acknowledgements and licences
Cardew’s speech recognition is built on open models and software. This page credits the people who made them and gives the licence terms we use them under.
Speech recognition model
NVIDIA Parakeet TDT 0.6B v3.
- Creator: NVIDIA Corporation.
- Original model: huggingface.co/nvidia/parakeet-tdt-0.6b-v3
- Licence: Creative Commons Attribution 4.0 International (CC BY 4.0)
- Core ML conversion: by FluidInference, huggingface.co/FluidInference/parakeet-tdt-0.6b-v3-coreml.
- Changes from the original: converted to Apple’s Core ML format by FluidInference, with selected precision and quantisation. Greythorn Studio Works made no further changes to the model.
The model is provided by its licensors “as is”, without warranties of any kind, as set out in section 5 of the CC BY 4.0 licence. Nothing on this page adds terms or restrictions to the model; you may use it under the CC BY 4.0 licence yourself.
FluidAudio
- Creator: FluidInference. Source: github.com/FluidInference/FluidAudio.
- Licence: Apache License 2.0. The full text ships with the app.
- Changes: none. Cardew uses it unmodified to run the model on Apple hardware.
Other components
FluidAudio includes BSD-licensed clustering code (fastcluster), whose notice ships with the app. Apple frameworks are used through the macOS SDK and aren’t redistributed. Full third-party notices also ship with the download.
Trade marks and endorsement
NVIDIA and Parakeet are trademarks of NVIDIA Corporation. Cardew is not affiliated with or endorsed by NVIDIA or FluidInference. “Cardew” is the name of a product by Greythorn Studio Works.