File size: 5,674 Bytes

---
license: apache-2.0
tags:
- speech-enhancement
- denoising
- mlx
- apple-silicon
- deepfilternet
- safetensors
base_model: Rikorose/DeepFilterNet3
library_name: mlx
pipeline_tag: audio-to-audio
---

# DeepFilterNet3 — MLX

Real-time speech enhancement for Apple Silicon. Removes background noise from
speech audio. FP32 MLX weights converted from the official DeepFilterNet3
checkpoint, with BatchNorm fused into the convolutions.

- **2.1M params**, fp32 safetensors, **8.1 MB**
- 48 kHz native, 10 ms frames, 2-frame (20 ms) lookahead
- Network-only export: STFT, ERB feature extraction, and deep-filter
  application live in the runtime (`auxiliary.npz` ships the exact DSP
  constants)

## Model

| Detail | Value |
|--------|-------|
| Architecture | DeepFilterNet3 (encoder + ERB decoder + deep-filter decoder) |
| Parameters | 2,131,824 (BatchNorm fused, GRU biases folded) |
| Precision | float32 |
| Sample rate | 48 kHz (FFT 960, hop 480) |
| ERB bands / DF bins | 32 / 96 |
| Deep-filter order | 5, lookahead 2 frames |

## Files

| File | Size | Description |
|------|------|-------------|
| `model.safetensors` | 8.1 MB | fp32 network weights in MLX layouts |
| `auxiliary.npz` | 126 KB | ERB filterbank + inverse, Vorbis window, normalization init states |
| `config.json` | 1 KB | Model and DSP hyperparameters |
| `dfn3_mlx.py` | 9 KB | Pure-MLX reference implementation of the network |

## Quality

30 VoiceBank-DEMAND test clips (16 kHz mirror, resampled to 48 kHz for the
model); the neural network forward runs on MLX while STFT / ERB / deep-filter
post-processing stay in the upstream PyTorch pipeline — the same methodology
used for the [CoreML variant](https://huggingface.co/aufklarer/DeepFilterNet3-CoreML).

| Backend | PESQ | STOI | SI-SDR |
|---------|------|------|--------|
| Noisy input | 2.205 | 0.932 | 9.28 |
| PyTorch FP32 (reference) | 2.900 | 0.947 | 18.19 |
| **MLX, CPU stream** | **2.900** | **0.947** | **18.19** |
| **MLX, GPU stream** | **2.902** | **0.947** | **18.18** |

Network outputs match PyTorch within 9e-7 on the MLX CPU stream —
numerically exact for fp32. The GPU stream uses Metal fast-math
transcendentals (max output delta ~4e-3) with no measurable metric impact.

## Latency (Apple M5 Pro, network forward only, Python MLX)

| Duration | GPU | RTF | CPU | RTF |
|----------|-----|-----|-----|-----|
| 5 s | 0.14 s | 0.028 | 0.18 s | 0.036 |
| 10 s | 0.29 s | 0.029 | 0.40 s | 0.039 |
| 20 s | 0.54 s | 0.027 | 0.79 s | 0.040 |

## Usage

```python
import sys

import mlx.core as mx
from huggingface_hub import snapshot_download

model_dir = snapshot_download("aufklarer/DeepFilterNet3-MLX")
sys.path.append(model_dir)
from dfn3_mlx import DFN3MLX

model = DFN3MLX(model_dir)

# Normalized features from your DSP front-end (constants in auxiliary.npz):
feat_erb = mx.zeros((1, 100, 32, 1))   # ERB features, dB-scaled + mean-normalized
feat_spec = mx.zeros((1, 100, 96, 2))  # complex spectrum (real, imag), unit-normalized

erb_mask, df_coefs, lsnr = model(feat_erb, feat_spec)
```

Full audio-in/audio-out enhancement additionally needs the DeepFilterNet DSP
path (STFT → features → ERB mask + deep filtering → iSTFT), documented in the
[upstream repository](https://github.com/Rikorose/DeepFilterNet).

On Apple devices, [speech-swift](https://github.com/soniqo/speech-swift)
runs this model directly via its MLX engine (or use the
[CoreML variant](https://huggingface.co/aufklarer/DeepFilterNet3-CoreML) for
the Neural Engine):

```swift
import SpeechEnhancement

let enhancer = try await SpeechEnhancer.fromPretrained(engine: .mlx)
let clean = try enhancer.enhance(audio: noisyAudio, sampleRate: 48000)
```

CLI:

```bash
swift run speech denoise noisy.wav --engine mlx
```

## Tensor contract

Channels-last `[B, T, F, C]` — time is the conv H axis, frequency the W axis.
The 2-frame lookahead shift is applied to the inputs inside the model.

| Direction | Name | Shape | Notes |
|-----------|------|-------|-------|
| input | `feat_erb` | `[B, T, 32, 1]` | ERB features, dB-scaled, exp-mean-normalized |
| input | `feat_spec` | `[B, T, 96, 2]` | complex spectrum features, unit-normalized |
| output | `erb_mask` | `[B, T, 32, 1]` | sigmoid ERB gain mask |
| output | `df_coefs` | `[B, 5, T, 96, 2]` | deep-filter coefficients (order, real/imag) |
| output | `lsnr` | `[B, T, 1]` | local SNR estimate, dB in [-15, 35] |

## Weight conventions

- **Conv2d**: `[O, kH, kW, I/groups]` (PyTorch `[O, I/g, kH, kW]` transposed);
  causal time padding, symmetric frequency padding
- **Depthwise ConvTranspose2d** (`erb_dec.convt{1,2}.dwt`): `[C, kH, kW, 1]`,
  true transposed-conv kernels (not pre-flipped); k=(1,3), freq stride 2,
  padding 1, output padding 1
- **GRU** (`mlx.nn.GRU` convention, gate order r/z/n): `Wx = weight_ih`,
  `Wh = weight_hh`, `b = bias_ih + [bias_hh_r; bias_hh_z; 0]`,
  `bhn = bias_hh_n`; run with an explicit zero initial hidden state
- **Grouped linear**: `[groups, in/groups, out/groups]`, applied as
  `einsum("btgi,gih->btgh")` then flattened
- BatchNorm is fused into the preceding conv; the unused `df_fc_a` head is
  dropped

## Source

- Base model: [Rikorose/DeepFilterNet](https://github.com/Rikorose/DeepFilterNet)
  v0.5.6 checkpoint (Apache-2.0 / MIT)
- Paper: [DeepFilterNet3](https://arxiv.org/abs/2305.08227)

## License

- Model weights: Apache-2.0 / MIT dual license
- MLX conversion: Apache-2.0

## Links

- [speech-swift](https://github.com/soniqo/speech-swift) — Apple SDK
- [Docs](https://soniqo.audio/getting-started) — install and CLI docs
- [soniqo.audio](https://soniqo.audio) — website
- [blog](https://soniqo.audio/blog) — blog