File size: 2,558 Bytes
d1e3119
 
 
e7a4ffe
 
086f35d
d1e3119
086f35d
 
 
 
e7a4ffe
 
 
 
e49f300
76324d4
e49f300
 
76324d4
e7a4ffe
4c769d9
76324d4
 
e7a4ffe
 
 
 
 
 
 
2d0b945
 
e7a4ffe
2d0b945
 
e7a4ffe
 
2d0b945
 
e7a4ffe
76324d4
 
 
 
bfa0c6f
 
 
76324d4
e7a4ffe
 
 
76324d4
e7a4ffe
 
 
 
 
 
 
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
---
language:
- en
- zh
license: mit
pipeline_tag: automatic-speech-recognition
tags:
- ASR
- Transcriptoin
- Diarization
- Speech-to-Text
library_name: transformers
---


## VibeVoice-ASR
[![GitHub](https://img.shields.io/badge/GitHub-Repo-black?logo=github)](https://github.com/microsoft/VibeVoice)
[![Live Playground](https://img.shields.io/badge/Live-Playground-green?logo=gradio)](https://aka.ms/vibevoice-asr)

**VibeVoice-ASR** is a unified speech-to-text model designed to handle **60-minute long-form audio** in a single pass, generating structured transcriptions containing **Who (Speaker), When (Timestamps), and What (Content)**, with support for **Customized Hotwords**.

➡️ **Code:** [microsoft/VibeVoice](https://github.com/microsoft/VibeVoice)<br>
➡️ **Demo:** [VibeVoice-ASR-Demo](https://aka.ms/vibevoice-asr)

<p align="left">
  <img src="figures/VibeVoice_ASR_archi.png" alt="VibeVoice-ASR Architecture" height="250px">
</p>


## 🔥 Key Features

- **🕒 60-minute Single-Pass Processing**:
  Unlike conventional ASR models that slice audio into short chunks (often losing global context), VibeVoice ASR accepts up to **60 minutes** of continuous audio input within 64K token length. This ensures consistent speaker tracking and semantic coherence across the entire hour.

- **👤 Customized Hotwords**:
  Users can provide customized hotwords (e.g., specific names, technical terms, or background info) to guide the recognition process, significantly improving accuracy on domain-specific content.

- **📝 Rich Transcription (Who, When, What)**:
  The model jointly performs ASR, diarization, and timestamping, producing a structured output that indicates *who* said *what* and *when*.




## Evaluation
<p align="center">
  <img src="figures/DER.jpg" alt="DER" width="70%">
  <img src="figures/cpWER.jpg" alt="cpWER" width="70%">
  <img src="figures/tcpWER.jpg" alt="tcpWER" width="70%">
</p>

## Installation and Usage

Please refer to [GitHub README](https://github.com/microsoft/VibeVoice/blob/main/docs/vibevoice-asr.md#installation).

## License
This project is licensed under the MIT License.

## Contact
This project was conducted by members of Microsoft Research. We welcome feedback and collaboration from our audience. If you have suggestions, questions, or observe unexpected/offensive behavior in our technology, please contact us at VibeVoice@microsoft.com.
If the team receives reports of undesired behavior or identifies issues independently, we will update this repository with appropriate mitigations.