README.md

---
dataset_info:
  features:
  - name: audio
    dtype: audio
  - name: text
    dtype: string
  - name: speaker
    dtype: string
  splits:
  - name: train
    num_bytes: 401319194100.855
    num_examples: 1042919
  - name: validation
    num_bytes: 52676843.0
    num_examples: 141
  - name: test
    num_bytes: 142910421.0
    num_examples: 383
  download_size: 167165459930
  dataset_size: 401514781364.855
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
task_categories:
- text-to-speech
language:
- vi
pretty_name: PhoAudiobook
extra_gated_prompt: >
  User agrees 
  (1) to use PhoAudiobook for research or educational purposes only,
  (2) to not distribute PhoAudiobook or part of PhoAudiobook in any original or modified form,
  (3) and to cite our [ACL 2025 paper](https://arxiv.org/abs/2506.01322) whenever PhoAudiobook is employed to help produce published results.
---

# PhoAudiobook: A high-quality zero-shot TTS dataset for Vietnamese

PhoAudiobook is a high-quality and large-scale Vietnamese speech dataset curated for zero-shot text-to-speech. Details of the dataset construction and experimental results can be found in our [ACL 2025 paper](https://arxiv.org/abs/2506.01322):

```bibtex
@inproceedings{vu2025zeroshottexttospeechvietnamese,
      title={Zero-Shot Text-to-Speech for Vietnamese}, 
      author={Thi Vu and Linh The Nguyen and Dat Quoc Nguyen},
      year={2025},
      booktitle={Proceedings of ACL},
}
```

## Data splits

**Train**

- 940 hours of audio longer than 10s
- 554 hours of augmented short audio
- Total: 1,494 hours of audio
- 710 speakers

**Validation**

- 0.5 hours of audio
- 5 speakers

**Test**

- "Seen" Speaker Test Set:
  - 0.8 hours of audio
  - 20 speakers (part of training speakers)
- "Unseen" Speaker Test Set:
  - 0.4 hours of audio
  - 20 speakers 


## User agreement

User agrees:
-   to use the dataset for research or educational purposes only.
-   to not distribute or part of the dataset in any original or modified form.
-   and to cite our ACL 2025 paper "Zero-Shot Text-to-Speech for Vietnamese" whenever the dataset is used to help produce published results.

## Disclaimer

	THE DATA IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
	IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
	FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
	AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
	LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
	OUT OF OR IN CONNECTION WITH THE DATA OR THE USE OR OTHER DEALINGS IN THE
	DATA.