MediaMotionYour archive has been sitting still. Put it to work.
Broadcasters, radio groups, production studios and podcast networks own the most valuable speech data on the continent. We license it for AI training — on your terms, with your veto, and with every hour accounted for.
If people are talking in it, we want to see it.
The value is in natural speech across South Africa's languages — the more varied the speakers, settings and registers, the better.
Content we take
Formats & logistics
Three steps. No surprises in step four.
We assess, you decide
Send us a catalogue list or a sample drive. Within two weeks you get a written valuation, a rights review and a list of what we would and wouldn't ingest.
- No cost, no obligation
- Chain-of-title check on a sample
- Talent and music-cue flags raised early
A licence built for training
Non-exclusive by default. Use restricted to AI model development — never broadcast, resale or public display. Exclude any title, series or presenter you choose.
- Per-hour fee plus revenue share on downstream licences
- Term, territory and buyer categories negotiable
- Termination and takedown provisions included
We process, you get paid
We ingest, clean, transcribe and package. Every downstream licence is logged against your catalogue and reported quarterly with payment.
- Provenance ledger, per file
- Quarterly statements and payouts
- Audit rights written into the agreement
Everything that matters.
- Copyright and all distribution rights
- The right to license the same content elsewhere
- A veto over any title, series, presenter or clip
- Approval over buyer categories (e.g. no defence, no ad-tech)
- Attribution or anonymity in our catalogue — your call
The parts nobody has time for.
- Digitisation of legacy tape formats
- Rights review and clearance summaries
- PII redaction and POPIA compliance
- Transcription, translation and alignment
- Buyer due diligence, contracting and collection
The honest answers.
Will a model trained on our content be able to reproduce it?
Our licences prohibit buyers from training models whose purpose is to reproduce your programming, and from using outputs to compete with your distribution. Speech models learn how a language sounds, not how to replay a specific episode. We also fingerprint delivered files so we can detect misuse.
What about presenters, actors and guests?
We review talent agreements as part of the audit and flag anything that needs additional clearance. Where a presenter or performer's likeness or voice could be a concern, we can exclude them, anonymise, or arrange a separate consent — you choose. Nothing ships that fails the rights review.
Do we have to give you everything?
No. Pick titles, series, date ranges or even individual episodes. Many partners start with a single show to see how it goes.
How is pricing set?
A per-hour licence fee based on language mix, speaker count, audio quality and exclusivity, plus a share of every downstream licence we sell. We put the numbers in writing at audit stage so you can compare.
Who are the buyers?
Frontier AI labs, well-funded speech and video startups, and research institutions. You approve the buyer categories in the agreement, and every licence is reported to you by name where NDAs allow.
What if we want out?
Termination for convenience with notice is standard. Existing downstream licences run to their term; no new ones are issued. Undelivered material is deleted and confirmed in writing.
Start with one show. See the audit, see the numbers, see the contract. Then decide about the rest of the vault.— How most of our partnerships begin