File size: 3,969 Bytes
8809972
 
 
 
 
 
 
 
 
 
 
 
 
39113db
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ecda4a9
39113db
 
 
 
 
 
 
 
039db64
 
 
 
 
 
 
 
39113db
 
 
 
 
 
2b3cbf4
 
39113db
 
 
 
039db64
39113db
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: cc0-1.0
task_categories:
- text-to-speech
- automatic-speech-recognition
language:
- 'no'
- nb
tags:
- librivox
pretty_name: NB-LibriVox
size_categories:
- 1K<n<10K
---

# 📄 NB-LibriVox

A high-quality Norwegian text-to-speech (TTS) dataset derived from LibriVox public domain audiobooks. It includes audio clips with pseudo-aligned transcripts and punctuation, curated by the National Library of Norway for speech synthesis and ASR research.

---

## 📂 Dataset Overview

| **Field**         | **Description**                                                                 |
|-------------------|---------------------------------------------------------------------------------|
| `file_name`       | Audio file name in `.wav` format                                                |
| `id`              | Unique identifier for each utterance/sample                                     |
| `text`            | Transcript automatically generated using NB-Whisper Large and nb-wav2vec2       |
| `text_punct`      | Transcript with punctuation automatically added using Qwen3                     |
| `speaker_id`      | Identifier of the speaker                                                       |
| `speaker_name`    | Name of the speaker                                                             |
| `speaker_gender`  | Gender of the speaker (e.g., `male`, `female`, `unknown`)                       |
| `speaker_status`  | Speaker's native status (e.g., `L1`, `L2`)                                      |
| `title`           | Title of the source audiobook                                                   |
| `author`          | Author of the original text                                                     |
| `sequence`        | Position of the utterance within the audiobook file                             |
| `duration`        | Duration of the audio clip in seconds                                           |

### Speaker Stats by Sample Count

| speaker_name     | speaker_id           | speaker_status | sample_count | total_duration (hrs) | avg_duration (secs) |
|------------------|----------------------|---------------|---------------|-----------------------|----------------------|
| Kathrine Engan   | female_no_librivox_1 | L1 | 4,873         | 9.0871                | 6.713                |
| Kristin G        | female_no_librivox_2 | L1 | 18            | 0.0525                | 10.388               |
| Austin Smith     | male_no_librivox_male_1 | L2 | 16            | 0.0170                | 3.822                |

---

## 📌 Additional Information

- **Language**: Norwegian (Bokmål)
- **Audio Format**: MPEG ADTS, layer III, v1, 64 kbps, 48 kHz, Monaural
- **Text Alignment**:  
  - **Method**: Pseudo-labeling and splitting of sentences up to 30 seconds
  - **ASR Model**: [`NB-Whisper Large`](https://huggingface.co/NbAiLab/nb-whisper-large)  
  - **Supplementary Model**: [`nb-wav2vec2`](https://huggingface.co/NbAiLab/nb-wav2vec2-1b-bokmaal-v2)  
  - **Alignment Type**: Pseudo-labeling (not force-aligned)
- **Punctuation Model**: [Qwen3 235B](https://huggingface.co/Qwen/Qwen3-235B-A22B)
- **Total Duration**: 9 hours
- **Total Samples**: 4907 samples

---

## 📜 Licensing & Usage

- **License**: Creative Commons Zero (CC0)  
- **Data Source**: LibriVox public domain recordings  
- **Dataset Creator**: National Library of Norway

> Free to use, modify, and distribute for any purpose, including commercial use.


---

## ⚠️ Ethical Considerations

- All content is in the public domain and CC0-licensed.
- Speakers are volunteers; synthesized voices should not be used in misleading or impersonating contexts.
- Encourage transparency when using in generative applications (e.g., voice assistants, synthetic speech).

---

## 📎 Citation

> “Norwegian LibriVox TTS Dataset (CC0), curated by the National Library of Norway. Transcripts pseudo-labeled using NB-Whisper Large and nb-wav2vec2.”