File size: 2,529 Bytes
2bd9217 dbab2e5 2bd9217 a20ad6c 2bd9217 7b3cf26 2bd9217 7b3cf26 2bd9217 6cce464 3e7a8bb 7b3cf26 3e7a8bb |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 |
---
license: apache-2.0
pipeline_tag: feature-extraction
library_name: mlx
tags:
- audio
- audio-to-audio
- codec
---
# NanoCodec for Apple Silicon
This is an MLX implementation of [NVIDIA NeMo NanoCodec](https://huggingface.co/nvidia/nemo-nano-codec-22khz-0.6kbps-12.5fps), a lightweight neural audio codec.
## Model Description
- **Architecture**: fully convolutional generator neural network and three discriminators. The generator comprises an encoder, followed by vector quantization, and a [HiFi-GAN-based](https://arxiv.org/abs/2010.05646) decoder.
- **Sample Rate**: 22.05 kHz
- **Framework**: MLX
- **Parameters**: 105M
## Installation
```bash
pip install nanocodec-mlx soundfile
```
## Usage
```python
from nanocodec_mlx.models.audio_codec import AudioCodecModel
import soundfile as sf
import mlx.core as mx
import numpy as np
# Load model from HuggingFace Hub
model = AudioCodecModel.from_pretrained("nineninesix/nemo-nano-codec-22khz-0.6kbps-12.5fps-MLX")
# Load audio
audio, sr = sf.read("input.wav")
audio_mlx = mx.array(audio, dtype=mx.float32)[None, None, :]
audio_len = mx.array([len(audio)], dtype=mx.int32)
# Encode and decode
tokens, tokens_len = model.encode(audio_mlx, audio_len)
reconstructed, recon_len = model.decode(tokens, tokens_len)
# Save output
output = np.array(reconstructed[0, 0, :int(recon_len[0])])
sf.write("output.wav", output, 22050)
```
#### Input
- **Input Type:** Audio
- **Input Format(s):** .wav files
- **Input Parameters:** One-Dimensional (1D)
- **Other Properties Related to Input:** 22050 Hz Mono-channel Audio
#### Output
- **Output Type**: Audio
- **Output Format:** .wav files
- **Output Parameters:** One Dimensional (1D)
- **Other Properties Related to Output:** 22050 Hz Mono-channel Audio
## License
This code is licensed under the Apache License 2.0.
The original NVIDIA NeMo NanoCodec model weights and architecture are developed by NVIDIA and are licensed under the [NVIDIA Open Model License](https://developer.download.nvidia.com/licenses/nvidia-open-model-license-agreement-june-2024.pdf). See [NOTICE](NOTICE) for attribution.
When using this project, you must comply with both licenses.
## Citation
This is an MLX implementation of NVIDIA NeMo NanoCodec. If you use this work, please cite the original:
- [NVIDIA NeMo NanoCodec](https://huggingface.co/nvidia/nemo-nano-codec-22khz-0.6kbps-12.5fps)
- [NVIDIA Open Model License](https://developer.download.nvidia.com/licenses/nvidia-open-model-license-agreement-june-2024.pdf) |