Tiv speech and text data · Available now
Studio-recorded Tiv speech and text data for TTS, ASR and translation.
Verified by native speakers. Recorded with documented consent. Built for teams training voice and language models in underrepresented Nigerian languages.
Listen
Hear the data.
Three clips from the Tiv corpus, each paired with its verified text and English translation.
Audio loads only when you press play.
[TODO: Tiv sentence for sample-01]
[TODO: English translation for sample-01]
[TODO: Audio pending: add sample-01.mp3 to /public/audio]
Why teams use it
Studio-recorded, not scraped or synthetic.
Native speakers read verified text in a controlled studio setup. No web scraping. No synthetic voices.
Verified by native speakers.
Sentences are translated and reviewed by native Tiv linguists, across dialect variation.
Consent documented for every speaker.
Each speaker signs a consent and release form. Each clip carries metadata, so you can audit provenance.
Tiv is tonal. Models cannot learn it from text alone.
What is available
A parallel corpus, with studio audio released category by category.
- Language
- Tiv, with English translations
- Text corpus
- 7,000+ parallel sentences
- Categories
- 27, from healthcare and government to idioms and proverbs
- Audio
- Studio-recorded, one clip per sentence, released category by category
- Uses
- Text-to-speech, speech recognition, machine translation, evaluation
- Licence
- Commercial, non-exclusive. Terms on request
- Open samples
- CC BY-NC 4.0, for non-commercial evaluation
Tell us what you are building.
Tell us which language you need and what it is for. Text-to-speech, speech recognition or translation.