Hugging Face's logo

stepfun-ai
/
Step-Audio-Chat

Audio-Text-to-Text
Safetensors
step1
custom_code
Model card Files Community
10
Step-Audio-Chat
File size: 9,719 Bytes 
40bdebd
 
e0335e2
40bdebd
39c2089
cf583b0
39c2089
7c5e63b
39c2089
cf583b0
 
 
 
 
e719f93
 
cf583b0
 
 
 
e719f93
 
cf583b0
 
 
 
e719f93
 
 
cf583b0
 
 
 
e719f93
 
cf583b0
 
 
 
e719f93
cf583b0
 
 
 
7c5e63b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
39c2089
7c5e63b
c9dde62
 
cf583b0
7c5e63b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cf583b0
7c5e63b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c9dde62
 
cf583b0
c9dde62
 
 
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
 
---
license: apache-2.0
pipeline_tag: audio-text-to-text
---

## 1. Step-Audio-Chat

This repository contains the Multimodal Large Language Model (LLM) component of Step-Audio. It is a 130 billion parameter multimodal LLM that is responsible for understanding and generating human speech. The model is specifically designed to seamlessly integrate functions such as speech recognition, semantic understanding, dialogue management, voice cloning, and speech generation.


## 2. Examples
### Clone audio
| role   | prompt wav | clone wav |
|:-------:|:-------:|:-------:|
|于谦| <audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/prompt_wav_yuqian.wav"></audio>|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/clone_wav_yuqian.wav"></audio>|
|李雪琴| <audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/prompt_wav_lixueqin.wav"></audio>|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/clone_wav_lixueqin.wav"></audio>|

### Speed control
| prompt | response |
|:-------:|:-------:|
|Human: 说一个绕口令<br>Assistant: 吃葡萄不吐葡萄皮，不吃葡萄倒吐葡萄皮<br>Human: 哎，你能把这个绕口令说的再快一点吗？|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/speed_control1.wav"></audio>|
|Human: 说一个绕口令<br>Assistant: 吃葡萄不吐葡萄皮，不吃葡萄倒吐葡萄皮<br>Human: 哎，你能把这个绕口令说的再快一点吗？<br>Assistant: 吃葡萄不吐葡萄皮，不吃葡萄倒吐葡萄皮<br>Human: 呃，你再用非常非常慢的速度说一遍的。|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/speed_control2.wav"></audio>|

### High EQ(emotional control & tone control)
| prompt | response |
|:-------:|:-------:|
|Human: 你这语气又不撒娇又不卖萌的，要不你撒个娇卖个萌吧。|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/tone_control.wav"></audio>|
|Human: 怎么办？我感觉我的人生很失败。|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/emotional_control1.wav"></audio>|
|Human: 小跃。你真的是。特别厉害。|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/emotional_control2.wav"></audio>|

### Multilingual (e.g., Chinese, English, Japanese)
| prompt | response |
|:-------:|:-------:|
|Human: What did the speaker mean when they said, it's raining cats and dogs?<br>Assistant: When they say "It's raining cats and dogs," it just means it's raining really hard. The speaker isn't literally saying cats and dogs are falling from the sky! It's just a fun way to describe heavy rain.|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/multilingual2.wav"></audio>|
|Human: こんにちは。（你好）<br>Assistant：こんにちは！何か手伝いましょうか？（您好！我可以帮你做点什么吗？）|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/multilingual1.wav"></audio>|

### Rap & Vocal
| prompt | response |
|:-------:|:-------:|
|human：唱一段rap|<audio controls src="https://huggingface.co/stepfun-ai/Step-Audio-Examples/resolve/main/rap.wav"></audio>|


## 3. Evaluation
### 3.1 LLM judge metrics(GPT-4o) on [**StepEval-Audio-360**](https://huggingface.co/datasets/stepfun-ai/StepEval-Audio-360)
<table>
    <caption>Comparison of fundamental capabilities of voice chat on the StepEval-Audio-360.</caption>
    <thead>
        <tr>
            <th>Model</th>
            <th style="text-align:center">Factuality (% &uarr;)</th>
            <th style="text-align:center">Relevance (% &uarr;)</th>
            <th style="text-align:center">Chat Score &uarr;</th>
        </tr>
    </thead>
    <tbody>
        <tr>
            <td>GLM4-Voice</td>
            <td style="text-align:center">54.7</td>
            <td style="text-align:center">66.4</td>
            <td style="text-align:center">3.49</td>
        </tr>
        <tr>
            <td>Qwen2-Audio</td>
            <td style="text-align:center">22.6</td>
            <td style="text-align:center">26.3</td>
            <td style="text-align:center">2.27</td>
        </tr>
        <tr>
            <td>Moshi<sup>*</sup></td>
            <td style="text-align:center">1.0</td>
            <td style="text-align:center">0</td>
            <td style="text-align:center">1.49</td>
        </tr>
        <tr>
            <td><strong>Step-Audio-Chat</strong></td>
            <td style="text-align:center"><strong>66.4</strong></td>
            <td style="text-align:center"><strong>75.2</strong></td>
            <td style="text-align:center"><strong>4.11</strong></td>
        </tr>
    </tbody>
</table>

*Note: Moshi are marked with "\*" and should be considered for reference only.


### 3.2 Public Test Set

<table>
    <thead>
        <tr>
            <th>Model</th>
            <th style="text-align:center">Llama Question</th>
            <th style="text-align:center">Web Questions</th>
            <th style="text-align:center">TriviaQA*</th>
            <th style="text-align:center">ComplexBench</th>
            <th style="text-align:center">HSK-6</th>
        </tr>
    </thead>
    <tbody>
        <tr>
            <td>GLM4-Voice</td>
            <td style="text-align:center">64.7</td>
            <td style="text-align:center">32.2</td>
            <td style="text-align:center">39.1</td>
            <td style="text-align:center">66.0</td>
            <td style="text-align:center">74.0</td>
        </tr>
        <tr>
            <td>Moshi</td>
            <td style="text-align:center">62.3</td>
            <td style="text-align:center">26.6</td>
            <td style="text-align:center">22.8</td>
            <td style="text-align:center">-</td>
            <td style="text-align:center">-</td>
        </tr>
        <tr>
            <td>Freeze-Omni</td>
            <td style="text-align:center">72.0</td>
            <td style="text-align:center">44.7</td>
            <td style="text-align:center">53.9</td>
            <td style="text-align:center">-</td>
            <td style="text-align:center">-</td>
        </tr>
        <tr>
            <td>LUCY</td>
            <td style="text-align:center">59.7</td>
            <td style="text-align:center">29.3</td>
            <td style="text-align:center">27.0</td>
            <td style="text-align:center">-</td>
            <td style="text-align:center">-</td>
        </tr>
        <tr>
            <td>MinMo</td>
            <td style="text-align:center">78.9</td>
            <td style="text-align:center">55.0</td>
            <td style="text-align:center">48.3</td>
            <td style="text-align:center">-</td>
            <td style="text-align:center">-</td>
        </tr>
        <tr>
            <td>Qwen2-Audio</td>
            <td style="text-align:center">52.0</td>
            <td style="text-align:center">27.0</td>
            <td style="text-align:center">37.3</td>
            <td style="text-align:center">54.0</td>
            <td style="text-align:center">-</td>
        </tr>
        <tr>
            <td><strong>Step-Audio-Chat</strong></td>
            <td style="text-align:center"><strong><i>81.0</i></strong></td>
            <td style="text-align:center"><strong>75.1</strong></td>
            <td style="text-align:center"><strong>58.0</strong></td>
            <td style="text-align:center"><strong>74.0</strong></td>
            <td style="text-align:center"><strong>86.0</strong></td>
        </tr>
    </tbody>
</table>

*Note: Results marked with "\*" on TriviaQA dataset are considered for reference only.*

*TriviaQA dataset marked with "\*" indicates results are for reference only.*

### 3.3 Audio instruction following
<table>
    <thead>
        <tr>
            <th rowspan="2">Category</th>
            <th colspan="2" style="text-align:center">Instruction Following</th>
            <th colspan="2" style="text-align:center">Audio Quality</th>
        </tr>
        <tr>
            <th style="text-align:center">GLM-4-Voice</th>
            <th style="text-align:center">Step-Audio</th>
            <th style="text-align:center">GLM-4-Voice</th>
            <th style="text-align:center">Step-Audio</th>
        </tr>
    </thead>
    <tbody>
        <tr>
            <td>Languages</td>
            <td style="text-align:center">1.9</td>
            <td style="text-align:center">3.8</td>
            <td style="text-align:center">2.9</td>
            <td style="text-align:center">3.3</td>
        </tr>
        <tr>
            <td>Role-playing</td>
            <td style="text-align:center">3.8</td>
            <td style="text-align:center">4.2</td>
            <td style="text-align:center">3.2</td>
            <td style="text-align:center">3.6</td>
        </tr>
        <tr>
            <td>Singing / RAP</td>
            <td style="text-align:center">2.1</td>
            <td style="text-align:center">2.4</td>
            <td style="text-align:center">2.4</td>
            <td style="text-align:center">4</td>
        </tr>
        <tr>
            <td>Voice Control</td>
            <td style="text-align:center">3.6</td>
            <td style="text-align:center">4.4</td>
            <td style="text-align:center">3.3</td>
            <td style="text-align:center">4.1</td>
        </tr>
    </tbody>
</table>


## 4. More information

For more information, please refer to our repository: [Step-Audio](https://github.com/stepfun-ai/Step-Audio).