Casper015/mlx-asr-benchmark
1 0PythonMITUpdated 1 month ago
Apple Silicon MLX ASR Benchmark & Local Transcription Suite: Empirically audited evaluation of 19 native MLX speech models on Apple Silicon (M5, 24GB Unified Memory)
apple-siliconasrmlxqwen3-asrspeech-recognitionsttwhisper
Our verdict: worth borrowing ideas from
Contains a compact Python benchmark suite and model‑selection logic for Apple‑silicon native ASR models that can be reused to pick or evaluate local transcription engines
Filed under speech to text in our directory.