Apr 24
MLX acceleration for local Parakeet models
Please consider using Apple MLX to speed up local Parakeet transcription on Apple Silicon. Lower latency would make local dictation feel more instant while keeping the privacy and offline benefits of local models.
This would be especially valuable for:
- Faster startup and transcription- Better Apple Silicon optimization- Stronger local-first and offline workflows- Better experience for heavy daily users- A more native, responsive feel on Mac
Inspired by Apple’s user experience philosophy, the most magical improvements often come from tiny interaction gains that make the whole product feel effortless.
Pending
Argmax is another options: https://github.com/argmaxinc/argmax-oss-swift