The acoustic model behind this service is published under a licence that asks for credit. This is that credit.
Speech recognition by NVIDIA Parakeet TDT 0.6B v3, © NVIDIA Corporation, used under CC-BY-4.0, which is offered without warranties. Modified: the published checkpoint has been converted to ONNX for inference, and quantised to int8 where a smaller model is served. The weights are otherwise unchanged; there has been no retraining or fine-tuning.
| Model | Licence | Commercial use | Asks of us |
|---|---|---|---|
nvidia/parakeet-tdt-0.6b-v3 | CC-BY-4.0 | Permitted, stated explicitly | Credit, a link, and a note of changes |
nvidia/parakeet-ctc-0.6b | CC-BY-4.0 | Permitted | The same |
openai/whisper-* | MIT | Permitted | Keep the copyright notice |
The service is built on open-source libraries, used unmodified. Their licences ask for notices to travel with copies of the software; this is a hosted service, and what leaves the building is a transcript, never a copy of the program, so no further notice is owed here.
A complete inventory is kept, and customers who need one for their own records are welcome to ask.