Feature Request: Great new VAD and ASR models

Author: D3vShoaibCreated Apr 16, 2026Updated Sep 5, 2026

I’d like to suggest adding support for a newer VAD and ASR stack.

FireRedVAD is multilingual, ~97.5% F1, streaming, theoretically better than Silero VAD.

NVIDIA Parakeet is streaming, English only, better than parakeet v2/v3.

Would be useful to have these as optional backends alongside the current setup.

Source: huggingface/speech-to-speech