### Install VoxMLX

Source: https://github.com/awni/voxmlx/blob/main/README.md

Install the voxmlx package using pip.

```bash
pip install voxmlx
```

--------------------------------

### Convert, Quantize, and Upload Voxtral Weights

Source: https://github.com/awni/voxmlx/blob/main/README.md

Use the -q flag for quantization and --upload-repo to specify a HuggingFace repository for uploading the converted model.

```bash
voxmlx-convert -q --mlx-path voxtral-mlx-4bit --upload-repo username/voxtral-mlx-4bit
```

--------------------------------

### Transcribe Audio from File

Source: https://github.com/awni/voxmlx/blob/main/README.md

Use the --audio flag with the voxmlx command to specify a path to an audio file for transcription.

```bash
voxmlx --audio audio.flac
```

--------------------------------

### Stream Audio from Microphone

Source: https://github.com/awni/voxmlx/blob/main/README.md

Run the voxmlx command without arguments to stream audio from your microphone for real-time transcription.

```bash
voxmlx
```

--------------------------------

### Convert and Quantize Voxtral Weights

Source: https://github.com/awni/voxmlx/blob/main/README.md

Use the -q flag with voxmlx-convert for 4-bit quantization. Specify the output directory with --mlx-path.

```bash
voxmlx-convert -q --mlx-path voxtral-mlx-4bit
```

--------------------------------

### Transcribe Audio Using Python API

Source: https://github.com/awni/voxmlx/blob/main/README.md

Import the transcribe function from the voxmlx library to transcribe an audio file programmatically. The transcribed text is then printed.

```python
from voxmlx import transcribe

text = transcribe("audio.flac")
print(text)
```

--------------------------------

### Convert Voxtral Weights to MLX Format

Source: https://github.com/awni/voxmlx/blob/main/README.md

Use the voxmlx-convert command to convert Voxtral model weights to the MLX format. Specify the output directory with --mlx-path.

```bash
voxmlx-convert --mlx-path voxtral-mlx
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.