File size: 9,041 Bytes
15a2b5f 07288e3 cc9919d bfdbdc6 44c61fb cc9919d 97228a0 cc9919d 07288e3 cc9919d 47655ca cc9919d 47655ca cc9919d 15a2b5f |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 |
---
license: apache-2.0
---
<p align="center">📑 <a href="">Technical Report</a>｜📖<a href="https://xqacmer.github.io/Ming-Unitok-Audio.github.io">Project Page</a> ｜🤗 <a href="https://huggingface.co/inclusionAI/MingTok-Audio">Hugging Face</a>｜ 🤖 <a href="https://modelscope.cn/models/inclusionAI/MingTok-Audio">ModelScope</a>
## Key Features
- 🚀 **First Unified Continuous Speech Tokenizer:** the first continuous audio tokenizer to effectively integrate semantic and acoustic features, suitable for both understanding and generation tasks.
- 🎧 **High-Quality Reconstruction:** Achieve high-quality audio generation by modeling continuous features with a VAE, minimizing information loss and preserving intricate acoustic textures.
- 🌐 **Convolution-Free Efficiency:** Built on a pure causal transformer architecture, completely eliminating convolutional layers for superior efficiency and a simpler design.
## Installation
```
pip install -r requirements.txt
```
## Quick start
```python
import torch
import torchaudio
from audio_tokenizer.modeling_audio_vae import AudioVAE
model = AudioVAE.from_pretrained('inclusionAI/MingTok-Audio')
model = model.cuda()
model.eval()
waveform, sr = torchaudio.load('data/1089-134686-0000.flac', backend='soundfile')
sample = {'waveform': waveform.cuda(), 'waveform_length': torch.tensor([waveform.size(-1)]).cuda()}
with torch.no_grad():
with torch.autocast(device_type='cuda', dtype=torch.bfloat16):
latent, frame_num = model.encode_latent(**sample)
output_waveform = model.decode(latent)
torchaudio.save('./1089-134686-0000_reconstruct.wav', output_waveform.cpu()[0], sample_rate=16000)
```
## Performance
### Speech reconstruction performance
<table>
<caption>Speech reconstruction performance comparison on various audio benchmark datasets. The best results are in <strong>bold</strong>.</caption>
<thead>
<tr>
<th rowspan="2" align="left"><b>System</b></th>
<th rowspan="2" align="center"><b>FrameRate</b></th>
<th colspan="3" align="center"><b>SEED-ZH</b></th>
<th colspan="3" align="center"><b>SEED-EN</b></th>
</tr>
<tr>
<th align="center"><b>PESQ↑</b></th>
<th align="center"><b>SIM↑</b></th>
<th align="center"><b>STOI↑</b></th>
<th align="center"><b>PESQ↑</b></th>
<th align="center"><b>SIM↑</b></th>
<th align="center"><b>STOI↑</b></th>
</tr>
</thead>
<tbody>
<tr>
<td align="left">MiMo-Audio-Tokenizer</td>
<td align="center">25</td>
<td align="center">2.71</td>
<td align="center">0.89</td>
<td align="center">0.93</td>
<td align="center">2.43</td>
<td align="center">0.85</td>
<td align="center">0.92</td>
</tr>
<tr>
<td align="left">GLM4-Voice-Tokenizer</td>
<td align="center">12.5</td>
<td align="center">1.06</td>
<td align="center">0.33</td>
<td align="center">0.61</td>
<td align="center">1.05</td>
<td align="center">0.12</td>
<td align="center">0.60</td>
</tr>
<tr>
<td align="left">Baichuan-Audio-Tokenizer</td>
<td align="center">12.5</td>
<td align="center">1.84</td>
<td align="center">0.78</td>
<td align="center">0.86</td>
<td align="center">1.62</td>
<td align="center">0.69</td>
<td align="center">0.85</td>
</tr>
<tr>
<td align="left">XY-Tokenizer</td>
<td align="center">12.5</td>
<td align="center">2.27</td>
<td align="center">0.77</td>
<td align="center">0.90</td>
<td align="center">2.14</td>
<td align="center">0.82</td>
<td align="center">0.90</td>
</tr>
<tr>
<td align="left">Mimi</td>
<td align="center">75</td>
<td align="center">2.05</td>
<td align="center">0.73</td>
<td align="center">0.89</td>
<td align="center">2.01</td>
<td align="center">0.77</td>
<td align="center">0.89</td>
</tr>
<tr>
<td align="left">XCodec2.0</td>
<td align="center">50</td>
<td align="center">2.19</td>
<td align="center">0.80</td>
<td align="center">0.92</td>
<td align="center">2.37</td>
<td align="center">0.82</td>
<td align="center">0.93</td>
</tr>
<tr>
<td align="left">BigCodec</td>
<td align="center">80</td>
<td align="center">2.26</td>
<td align="center">0.81</td>
<td align="center">0.92</td>
<td align="center">2.22</td>
<td align="center">0.80</td>
<td align="center">0.91</td>
</tr>
<tr>
<td align="left"><strong>MingTok-Audio(ours)</td>
<td align="center">50</td>
<td align="center"><b>4.21</b></td>
<td align="center"><b>0.96</b></td>
<td align="center"><b>0.98</b></td>
<td align="center"><b>4.04</b></td>
<td align="center"><b>0.96</b></td>
<td align="center"><b>0.98</b></td>
</tr>
</tbody>
</table>
### The adaptation performance for downstream ASR tasks
<table>
<caption>Understanding ASR performance comparison on various audio benchmark datasets. The best results are in <strong>bold</strong>.</caption>
<thead>
<tr>
<th rowspan="2"><strong>Datasets</strong></th>
<th rowspan="2"><strong>Model</strong></th>
<th colspan="7"><strong>Performance</strong></th>
</tr>
<tr>
<th><strong>aishell2-ios</strong></th>
<th><strong>LS-clean</strong></th>
<th><strong>Hunan</strong></th>
<th><strong>Minnan</strong></th>
<th><strong>Guangyue</strong></th>
<th><strong>Chuanyu</strong></th>
<th><strong>Shanghai</strong></th>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="4"><strong>Understanding ASR</strong></td>
<td>Kimi-Audio</td>
<td><strong>2.56</td>
<td><strong>1.28</td>
<td>31.93</td>
<td>80.28</td>
<td>41.49</td>
<td>6.69</td>
<td>60.64</td>
</tr>
<tr>
<td>Qwen2.5 Omni</td>
<td>2.75</td>
<td>1.80</td>
<td>29.31</td>
<td>53.43</td>
<td>10.39</td>
<td>7.61</td>
<td>32.05</td>
</tr>
<tr>
<td>Qwen2 Audio</td>
<td>2.92</td>
<td>1.60</td>
<td>25.88</td>
<td>123.78</td>
<td>7.59</td>
<td>7.77</td>
<td>31.73</td>
</tr>
<tr>
<td><strong>Ming-UniAudio-16B-A3B(ours)</td>
<td>2.84</td>
<td>1.62</td>
<td><strong>9.80</strong></td>
<td><strong>16.50</strong></td>
<td><strong>5.51</strong></td>
<td><strong>5.46</strong></td>
<td><strong>14.65</strong></td>
</tr>
</tbody>
</table>
### The adaptation performance for downstream TTS tasks
<table>
<caption>Performance comparison on various audio benchmark datasets. The best results are in <strong>bold</strong>.</caption>
<thead>
<tr>
<th align="left"><b>Datasets</b></th>
<th align="left"><b>Model</b></th>
<th colspan="4" align="center"><b>Performance</b></th>
</tr>
<tr>
<th></th>
<th></th>
<th align="center"><b>Seed-zh WER(%)</b></th>
<th align="center"><b>Seed-zh SIM</b></th>
<th align="center"><b>Seed-en WER(%)</b></th>
<th align="center"><b>Seed-en SIM</b></th>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="5" align="left" style="vertical-align: middle;"><b>Generation</b></td>
<td align="left">Seed-TTS</td>
<td align="center">1.12</td>
<td align="center"><b>0.80</b></td>
<td align="center">2.25</td>
<td align="center"><b>0.76</b></td>
</tr>
<tr>
<td align="left">MiMo-Audio</td>
<td align="center">1.96</td>
<td align="center">-</td>
<td align="center">5.37</td>
<td align="center">-</td>
</tr>
<tr>
<td align="left">Qwen3-Omni-30B-A3B-Instruct</td>
<td align="center">1.07</td>
<td align="center">-</td>
<td align="center"><b>1.39</b></td>
<td align="center">-</td>
</tr>
<tr>
<td align="left">Ming-Omni-Lite</td>
<td align="center">1.69</td>
<td align="center">0.68</td>
<td align="center">4.31</td>
<td align="center">0.51</td>
</tr>
<tr>
<td align="left"><strong>Ming-UniAudio-16B-A3B(ours)</td>
<td align="center"><b>0.95</b></td>
<td align="center">0.70</td>
<td align="center">1.85</td>
<td align="center">0.58</td>
</tr>
</tbody>
</table>
## Acknowledgements
1. We borrowed a lot of code from [X-Codec-2.0](https://github.com/zhenye234/X-Codec-2.0.git) for tokenizer training.
2. We thank the OpenAI team for developing the [Whisper](https://github.com/openai/whisper) model and making its weights publicly available.
## License and Legal Disclaimer
This code repository is licensed under the [MIT License](./LICENSE), and the Legal Disclaimer is located in the [LEGAL.md file](./LEGAL.md) under the project's root directory.
## Citation
If you find our work helpful, feel free to give us a cite. |