---
library_name: mlx
tags:
  - mlx
  - speech-recognition
  - asr
  - canary
  - apple-silicon
license: cc-by-4.0
language:
  - en
  - bg
  - hr
  - cs
  - da
  - nl
  - et
  - fi
  - fr
  - de
  - el
  - hu
  - it
  - lv
  - lt
  - mt
  - pl
  - pt
  - ro
  - sk
  - sl
  - es
  - sv
  - ru
  - uk
---

# Canary MLX

NVIDIA Canary ASR model converted to MLX format for Apple Silicon.

## Usage

```bash
pip install canary-mlx
```

```python
from canary_mlx import load_model

model = load_model("qfuxa/canary-mlx")
result = model.transcribe("audio.wav", language="en")
print(result)
```

## Model Details

This model is a conversion of NVIDIA's Canary ASR model to Apple's MLX framework.

- **Architecture**: Conformer encoder + Transformer decoder
- **Parameters**: ~1B
- **Supported Languages**: 25 languages (see tags)

## Original Model

Based on NVIDIA NeMo Canary model. See [NVIDIA NeMo](https://github.com/NVIDIA/NeMo) for the original implementation.

## License

Model weights are released under CC-BY-4.0 license (same as original NVIDIA model).