Datasets:

ODYSSEYAILABS
/

odyssey-sa-voice-corpus-v0.1

Languages:

Size:

Tags:

african-languages

License:

Dataset card Files Files and versions

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

language: - en - zu - xh - set - afr license: cc-by-4.0 task_categories: - automatic-speech-recognition - text-to-speech tags: - south-african - codeswitch - multilingual - african-languages - asr - tts pretty_name: Odyssey SA Voice Corpus v0.1 size_categories: 10K<n<100K

Odyssey SA Voice Corpus v0.1 — Ethical South African Code-Switching Dataset

20 hours of consented, high-quality South African voice data:

700 h natural podcast clips (30-sec, transcribed, creator-permitted)
20 h scripted code-switch bursts (5–10 sec)
Full CSV metadata (speaker age/gender/accent/language mix)

Free 1-hour sampler below ↓
Full 720 h + custom LaaS annotation service → DM or Fiverr link in bio
Commercial/exclusive licenses available on request.

Perfect for inclusive ASR/TTS and multilingual African models.

Downloads last month: 7