Loading datasets...
Release Date: 11/4/2026
Format: FLAC, TSV
Size: 49.25 GB
Synthetic Dholuo (luo) speech for automatic speech recognition, created by CLEAR Global. It contains 657,877 utterances totalling about 775 hours at 24 kHz mono, published as a single train split. The sentences were written by Claude Sonnet 3.7, and the audio was synthesised by a YourTTS model fine-tuned on Dholuo Bible recordings (open.bible). The voice is a single male voice. Each clip ships with a normalised transcript, the original LLM sentence and its duration. Audio is FLAC (lossless re-encoding of the original 24 kHz WAV).
Licensing
Creative Commons Attribution Non Commercial 4.0 International (CC-BY-NC-4.0)
https://spdx.org/licenses/CC-BY-NC-4.0.htmlThe synthetic text was generated with Claude Sonnet 3.7. Use of the content is limited to non-commercial purposes and must comply with Anthropic's Terms of Service and Usage Policy, and must not be used to develop products or services that compete with Anthropic's offerings.
Commercial use, and any use that violates Anthropic's Terms of Service or Usage Policy.
Ethical Review
The TTS model was fine-tuned on the open.bible Dholuo recordings from BibleTTS-Luo, read by a single male narrator. The narrator is not identified in this release, and no personal data is included. The text is generated and does not contain personal information.
Intended Use
Research and development of automatic speech recognition for Dholuo, including training with real data.
This is the Dholuo synthetic speech from DeRenzi et al. (2025). Dholuo has very little transcribed speech available, so synthetic data adds training material that would otherwise not exist.
The sentences were generated with Claude Sonnet 3.7. The audio was synthesised by a YourTTS model fine-tuned by CLEAR Global on Dholuo Bible recordings from the BibleTTS dataset.
The linguist-evaluated text sample from the same project (1,600 sentences per language, with readability and adequacy ratings) is published as CLEAR-Global/Synthetic-Text, also on Mozilla Data Collective.
clips/train/ — 657,877 FLAC files, 24 kHz mono, 16-bit, named by the MD5 of the original file name.metadata.tsv — one row per clip with the columns path, split, text, original_text, duration.text is lowercased with punctuation removed and is the transcript to use for ASR training. original_text keeps the LLM's capitalisation and punctuation.
DeRenzi, B., Dixon, A., Farhi, M.A., Resch, C. (2025) Synthetic Voice Data for Automatic Speech Recognition in African Languages. arXiv:2507.17578.
@misc{derenzi2025synthetic,
title = {Synthetic Voice Data for Automatic Speech Recognition in African Languages},
author = {DeRenzi, Brian and Dixon, Anna and Farhi, Mohamed Aymane and Resch, Christian},
year = {2025},
eprint = {2507.17578},
archivePrefix = {arXiv},
primaryClass = {cs.CL},
url = {https://arxiv.org/abs/2507.17578}
}
This dataset is synthetic: it was not recorded from human speakers. The text was generated by a large language model and the audio by a text-to-speech model, so it may contain model-induced biases. The voice is male only. The work was supported by the Gates Foundation; the conclusions and opinions expressed are those of the authors alone and shall not be attributed to the Foundation.
Please check README.md for more information.