diff --git a/whisperx/transcribe.py b/whisperx/transcribe.py
index 6fff837..edd2764 100644
--- a/whisperx/transcribe.py
+++ b/whisperx/transcribe.py
@@ -81,6 +81,7 @@ def cli():
     args = parser.parse_args().__dict__
     model_name: str = args.pop("model")
     batch_size: int = args.pop("batch_size")
+    model_dir: str = args.pop("model_dir")
     output_dir: str = args.pop("output_dir")
     output_format: str = args.pop("output_format")
     device: str = args.pop("device")
@@ -166,7 +167,7 @@ def cli():
     results = []
     tmp_results = []
     # model = load_model(model_name, device=device, download_root=model_dir)
-    model = load_model(model_name, device=device, device_index=device_index, compute_type=compute_type, language=args['language'], asr_options=asr_options, vad_options={"vad_onset": vad_onset, "vad_offset": vad_offset}, task=task, threads=faster_whisper_threads)
+    model = load_model(model_name, device=device, device_index=device_index, download_root=model_dir, compute_type=compute_type, language=args['language'], asr_options=asr_options, vad_options={"vad_onset": vad_onset, "vad_offset": vad_offset}, task=task, threads=faster_whisper_threads)
 
     for audio_path in args.pop("audio"):
         audio = load_audio(audio_path)