nvidia

/

multitalker-parakeet-streaming-0.6b-v1

like 130

Follow

NVIDIA 67k

Hi.
We will be looking into non-English ASR support in the near future.
Note that Sortformer speaker diarization model is not limited to English.
Only multitalker Parakeet ASR model is limited to English.

@taejinp Thanks for the info :) Do you know if a "parakeet-streaming-0.6b" model (w/o speaker diarization) is planned? So like nvidia/parakeet-tdt-0.6b-v3 but optimized for streaming?

Try this : https://huggingface.co/nvidia/parakeet_realtime_eou_120m-v1

so basically if one instance of this model accepts audio of 2 speakers, whoever speaks first will be transcribed only till the end of audio?

Hi.
We will be looking into non-English ASR support in the near future.
Note that Sortformer speaker diarization model is not limited to English.
Only multitalker Parakeet ASR model is limited to English.

Hi @taejinp ,

Thanks for the update!

With the release of Nemotron-3.5-ASR-Streaming-0.6B, are we any closer to seeing a multilingual version of the multitalker-parakeet-streaming model? If so, is there any rough timeline or expectation you can share, even if it's just a broad estimate?

Also, thank you to you and the NVIDIA NeMo team for the excellent work. The progress on streaming ASR and diarization has been really impressive, and we're looking forward to what's coming next!