+class MLXWhisper(ASRBase):

+    """

+    Uses MPX Whisper library as the backend, optimized for Apple Silicon.

+    Models available: https://huggingface.co/collections/mlx-community/whisper-663256f9964fbb1177db93dc

+    Significantly faster than faster-whisper (without CUDA) on Apple M1. Model used by default: mlx-community/whisper-large-v3-mlx

+    """

+

+    sep = " "

+

+    def load_model(self, modelsize=None, model_dir=None):

+        from mlx_whisper import transcribe

+

+        if model_dir is not None:

+            logger.debug(f"Loading whisper model from model_dir {model_dir}. modelsize parameter is not used.")

+            model_size_or_path = model_dir

+        elif modelsize is not None:

+            logger.debug(f"Loading whisper model {modelsize}. You use mlx whisper, so make sure you use a mlx-compatible model.")

+            model_size_or_path = modelsize

+        elif modelsize == None:

+            logger.debug("No model size or path specified. Using mlx-community/whisper-large-v3-mlx.")

+            model_size_or_path = "mlx-community/whisper-large-v3-mlx"

+        

+        self.model_size_or_path = model_size_or_path

+        return transcribe

+    

+    def transcribe(self, audio, init_prompt=""):

+        segments = self.model(

+            audio,

+            language=self.original_language,

+            initial_prompt=init_prompt,

+            word_timestamps=True,

+            condition_on_previous_text=True,

+            path_or_hf_repo=self.model_size_or_path,

+            **self.transcribe_kargs

+        )

+        return segments.get("segments", [])

+

+

+    def ts_words(self, segments):

+        """

+        Extract timestamped words from transcription segments and skips words with high no-speech probability.

+        """

+        return [

+            (word["start"], word["end"], word["word"])

+            for segment in segments

+            for word in segment.get("words", [])

+            if segment.get("no_speech_prob", 0) <= 0.9

+        ]

+    

+    def segments_end_ts(self, res):

+        return [s['end'] for s in res]

+

+    def use_vad(self):

+        self.transcribe_kargs["vad_filter"] = True

+

+    def set_translate_task(self):

+        self.transcribe_kargs["task"] = "translate"

+    parser.add_argument('--backend', type=str, default="faster-whisper", choices=["faster-whisper", "whisper_timestamped", "mlx-whisper", "openai-api"],help='Load only this backend for Whisper processing.')

+        elif backend == "mlx-whisper":

+            asr_cls = MLXWhisper