Datasetpublic

scubavoice/ibibio-efik-speech-corpus-sample

Scuba Voice Dataset: Ibibio & Efik Sample (1 hour) Scuba is a voice data infrastructure company collecting, labeling, and licensing speech corpora for low-resource languages. This repository contains a 1-hour sample of labeled Ibibio and Efik speech data collected from native speakers in Akwa Ibom, Nigeria. This sample is provided for evaluation purposes only. For access to the full dataset (~200 hours across Ibibio and Efik), contact us at info@scubavoice.com.… See the full description on the dataset page: https://huggingface.co/datasets/scubavoice/ibibio-efik-speech-corpus-sample.

sourceHugging Faceotherupdated 3mo agoView on Hugging Face

1likes25downloads

Dataset Card

Scuba Voice Dataset: Ibibio & Efik Sample (1 hour)

Scuba is a voice data infrastructure company collecting, labeling, and licensing speech corpora for low-resource languages. This repository contains a 1-hour sample of labeled Ibibio and Efik speech data collected from native speakers in Akwa Ibom, Nigeria.

This sample is provided for evaluation purposes only. For access to the full dataset (~200 hours across Ibibio and Efik), contact us at info@scubavoice.com.

Languages

Field	Ibibio	Efik
ISO 639-3	`ibb`	`efi`
Region	Akwa Ibom State, Nigeria	Akwa Ibom & Cross River State, Nigeria
Language family	Niger-Congo, Cross River	Niger-Congo, Cross River
Speakers (estimated)	~7 million	~3 million

Both Ibibio and Efik are tonal languages. This dataset includes prompt-type labels that reflect the emotional register of each utterance, which is relevant for tone-sensitive Speech-to-Intent, ASR and TTS modeling.

Dataset Structure

scuba-ibibio-efik-sample/
├── audio/               # .webm audio files
├── metadata.csv         # Per-clip metadata
└── README.md

Metadata fields

Column	Description
`file_name`	Path to audio file
`recording_id`	Unique clip identifier
`language`	Language name
`iso_639_3`	ISO 639-3 language code
`english_translation`	English gloss of the prompt
`prompt_type`	Emotional register (neutral, serious, worried, etc.)
`speaker_id`	Anonymized speaker identifier
`gender`	Speaker gender (self-reported)
`age_range`	Speaker age bracket
`duration_seconds`	Clip duration
`license`	License type
`consent_status`	Consent collection method
`consent_version`	Consent form version

Sample Statistics

Metric	Value
Total duration	~1 hour
Languages	Ibibio (`ibb`), Efik (`efi`)
Full dataset	100+ hours, Ibibio + Efik

License

This sample is released under the [Scuba Sample Voice Data License](https://huggingface.co/datasets/scubavoice/ibibio-efik-speech-corpus-sample/resolve/main/LICENSE).

—Free for evaluation and non-commercial research
—Commercial use requires a separate license agreement
—Redistribution is not permitted

To license the full dataset: info@scubavoice.com

Full Dataset Access

This sample represents <1% of Scuba's current corpus. The full dataset includes:

—~200 hours of labeled speech
—Ibibio and Efik (ISO: ibb, efi)
—500+ speakers across gender and age groups
—Dialect coverage across South South Nigeria
—Conversational english prompt text with emotional register labels (neutral, serious, worried, etc.)

For licensing, research partnerships, or custom collection: info@scubavoice.com

Citation

If you use this data in published research, please cite:

@dataset{scuba2026ibibio_efik,
  title     = {Scuba Ibibio-Efik Voice Dataset},
  author    = {Scuba},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/scuba-ai/ibibio-efik-sample}
}

About Scuba

Scuba collects, labels, and licenses speech corpora for languages that AI has left behind. We work directly inside speaker communities to build the data infrastructure that makes voice AI possible for everyone.

Website: scubavoice.com Contact: info@scubavoice.com