LIGHTNING WHISPER MLX

Summary from the TLDR newsletter

Lightning Whisper MLX is an incredibly fast implementation of Whisper optimized for Apple Silicon. It features batched decoding for higher throughput, distilled models for faster decoding, and quantized models for faster memory movement. Lightning Whisper MLX is 10 times faster than Whisper CPP and four times faster than the current MLX Whisper implementation.