Blog

Beyond Whisper: shipping NVIDIA Parakeet in a local-first meeting scribe

How we roughly halved local transcription error rates by moving from whisper.cpp to NVIDIA Parakeet on sherpa-onnx — and what Whisper still does better.