281M and 235M param models too. https://github.com/facebookresearch/seamless_communication/blob/main/docs/m4t/on_device_README.md https://github.com/facebookre

281M and 235M param models too. https://github.com/facebookresearch/seamless_communication/blob/main/docs/m4t/on_device_README.md https://github.com/facebookresearch/seamless_communication/b... I don't really know how the metrics they list compare to whisper, I'm very curious if these are fast enough for realtime speech2text? I think whisper technically could but it was difficult to do or something like that?