Hugging Face's logo

keyvan-ai
/
wakecore

Audio Classification
German
English
wakecore
audio
speech
hotword
wake-word
voice
keyword-spotting
on-device
private
edge
Model card Files
xet
 Community
wakecore
File size: 4,076 Bytes
c1a368a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
---
language:
  - de
  - en
license: mit
tags:
  - audio
  - speech
  - hotword
  - wake-word
  - voice
  - keyword-spotting
  - on-device
  - private
  - edge
pipeline_tag: audio-classification
library_name: wakecore
---

# WakeCore

> **Open hotword detection for private voice systems.**
> Lokale Hotword-Erkennung für private KI-Systeme.

WakeCore is a cross-language SDK family for embedding always-listening
hotword ("wake-word") detection into your applications. Detection runs
entirely on-device — no cloud calls, no API keys, no telemetry — so you
stay in control of the microphone signal.

**Drop a `.wake` file into your app, point WakeCore at it, and you get
on-device hotword detection.** Custom `.wake` files for your own phrases
are produced by the WakeCore generator at <https://wakecore.de>.

## Quick start

### Python

```bash
pip install wakecore[audio]
wakecore install-engine
```

```python
from wakecore import Runtime, AudioCapture

with Runtime.load("hey_computer.wake") as rt, \
     AudioCapture(rt.sample_rate, rt.frame_length) as mic:
    for frame in mic:
        if rt.process(frame):
            print("wake!")
```

### Node.js

```bash
npm install wakecore
```

```javascript
const { Runtime } = require("wakecore");
const rt = Runtime.load("hey_computer.wake");
// feed PCM-16 mono frames of rt.bytesPerFrame bytes
rt.close();
```

### Rust

```toml
[dependencies]
wakecore = "0.1"
```

```rust
let mut rt = wakecore::Runtime::load("hey_computer.wake")?;
let frame = [0i16; 512];
if rt.process(&frame)? { println!("wake!"); }
```

## Supported platforms (8 OS families, 17 binary variants)

| Family            | Architectures                                     |
|-------------------|---------------------------------------------------|
| Linux             | x86_64, i386                                      |
| macOS             | x86_64, arm64                                     |
| Windows           | amd64, i686                                       |
| Raspberry Pi      | cortex-a7, cortex-a53, arm11                      |
| ARM Linux         | a9-neon (generic ARMv7+NEON)                      |
| Android           | x86, x86_64, armeabi-v7a, arm64-v8a               |
| iOS               | universal static                                  |
| watchOS           | universal static                                  |

## What's in this repository

| Path                              | Purpose                                        |
|-----------------------------------|------------------------------------------------|
| `hotwords/*.wake`                 | Ready-to-use sample hotword files              |
| `engine/<plat>/<arch>/<ver>/`     | Native engine binary per platform              |
| `engine/<plat>/<arch>/<ver>/SHA256SUMS` | Integrity manifest                       |

## The `.wake` file format

`.wake` files are sealed binary containers — opaque to user code. The
SDK simply hands the body to the bundled inference engine. New `.wake`
files are produced by the WakeCore generator service.

## Audio format

- Sample rate: **16,000 Hz**
- Channels: **mono**
- Sample format: **16-bit signed little-endian PCM**
- Frame size: **512 samples** per `process()` call (~32 ms)

## Why WakeCore

| | WakeCore | Cloud APIs | Other open SDKs |
|-|----------|------------|------------------|
| On-device  | ✓ | ✗ | ✓ |
| No API key | ✓ | ✗ | ✓ |
| Cross-language | ✓ Python · Node.js · C · C++ · C# · Rust | varies | usually single-language |
| Free tier | ✓ first hotword free | rare | training-yourself |
| German-aware | ✓ | rarely | rarely |
| `.wake` files portable across platforms | ✓ | n/a | varies |

## Tags

`audio` · `speech` · `hotword` · `wake-word` · `keyword-spotting` ·
`voice-assistant` · `private-voice` · `german` · `on-device` ·
`edge-ai` · `iot`

## Links

- GitHub:   <https://github.com/keyvan-ai/wakecore>
- Website:  <https://wakecore.de>
- Generator: <https://api.wakecore.de>
- License:  [MIT](https://github.com/keyvan-ai/wakecore/blob/main/LICENSE)