What we do, and how far each function goes.
Texadata operates as an aggregation and clearance layer, not an archive owner. The archives belong to the institutions we onboard; our work is proving title, clearing rights for machine learning use, and delivering to a buyer's standard.
Video that does not exist in the corpora.
Aerial and ground capture of environments that Western-sourced datasets do not contain: dense informal urban fabric, waterfront and stilt settlements, open-air markets, two-wheeler traffic, and ceremonial gatherings at real scale.
Uncrewed, low and mid altitude
1080p and 4K, up to 106 Mbps, including 10-bit HEVC. Embedded flight telemetry retained rather than stripped on transcode. Sustained single takes for continuous scene tracking.
Handheld with uncompressed sync audio
1080p59.94 with PCM 48 kHz stereo captured in-camera — not a transcoded proxy. Speech, music and room response usable as an audio asset in its own right.
Purpose-built to a buyer's spec
Egocentric and fixed-rig collection against a written brief: scene taxonomy, camera discipline, subject consent and volume agreed before capture begins.
Languages that production models cannot hold.
Nigerian English, Nigerian Pidgin and code-switched broadcast audio, recorded in real acoustic conditions rather than a booth. Natural turn-taking, overlap on hand-offs, telephone-band callers against studio-band hosts.
Broadcast archive
Radio back-catalogues and call-in programming from stations across Nigeria and francophone Central Africa. Decades of in-language talk, anglophone and francophone, never licensed to anyone.
Commissioned conversational
Purpose-collected Pidgin and Nigerian English, code-switched, channel-separated, transcribed, with full speaker metadata and consent covering commercial AI training.
The function that actually carries the value.
Both of the disqualifying failures in this market are documentation failures, not sourcing failures. Rights are established per asset, never asserted in aggregate, and the underlying instrument is producible for any single item on request.
Masters, metadata and records — to your pipeline.
Original containers without recompression, JSON sidecars per asset, and transfer by direct object-storage handoff, buyer-side bucket, or physical media for archive volumes.