Skip to content

Insights

Latest

From the Motion56 team.

  1. Audio deepfake detection datasets: why lab results don't hold

    Audio deepfake detection datasets decide whether a voice clone detector works on real calls and recordings. What ASVspoof and In-the-Wild show, and what real speech data should add.

  2. World model training data: what video world models learn from

    World model training data is mostly video: millions of hours of it, cut, filtered and labelled. What the best-known world models were trained on and why real footage still matters.

  3. Licensed vs scraped vs synthetic video: what makes great AI training data

    Scraped, synthetic or licensed? How the three main sources of video training data compare, and which one to build your model on.

  4. How to buy AI training data: the brief, the sample and the paperwork

    How to buy AI training data without paying for the wrong footage: what to put in the brief, how to test a sample and the documentation to ask for before you sign.

  5. Do AI video models understand physics?

    Do AI video models understand physics? The benchmarks say not yet. What Physics-IQ, VideoPhy and other tests found, where the models fail and what kind of data could help.

  6. Video archive digitisation: why tape can't wait

    Video archive digitisation decides whether footage on tape survives, and whether it can ever be licensed. Why archivists set 2025 as a deadline, what is at risk and what to do first.

  7. GDPR and AI training: what the rules say about people in video

    GDPR and AI training meet whenever people can be identified in the data. The lawful basis, when faces and voices count as biometric data, and what European and UK regulators have said.

  8. Multilingual speech datasets: why accents still trip up voice AI

    Multilingual speech datasets decide which voices AI understands. How much non-English audio big models use, how error rates change with accent and what a better speech set includes.

  9. Dashcam training data: why real crashes are so scarce

    Dashcam training data teaches driving and fleet safety models what a crash looks like before it happens. Why real crashes and near-misses are scarce, and what the open datasets hold.

  10. Open video datasets: which ones can you use commercially?

    Open video datasets are easy to download and hard to use commercially. What the licences of the best-known sets allow, what happened to WebVid and what it means for commercial teams.

  11. The role of scripted content in generative AI production

    Real-world footage teaches AI how life looks. Scripted content teaches it how a scene works. Here’s what that footage offers generative video, where it fits in production, and why performers’ rights have to come first.

  12. Synthetic vs real training data: where video AI breaks

    Synthetic vs real training data is one of the big questions in video AI. Where generated footage helps, where models still break on the real world and what mix the research supports.

Guides and events

For AI teams and content owners.

  1. How AI training data licensing works

    For AI teams: what a licence covers, how it is priced and how a brief becomes a licence.

  2. How to license your video archive for AI training

    For content owners: what happens, what it costs, what you keep and how you are paid.

  3. Out of the lab: what happens when video AI meets the real world

    An evening panel discussion in London on Thursday 19 November 2026.

Topics

Browse by topic.

  1. Content owners

    For archives, broadcasters and rights holders licensing video to AI.

  2. Licensing

    How AI training data is licensed and priced, and the law behind it.

  3. Metadata

    Making video searchable and licensable for AI training.

  4. Training data

    What AI teams need from real-world video, use by use.

Why Motion56

Volume. Accuracy. Speed. Trust.

  • Volume

    79,133 hours

    Of real-world video across our categories, and sourced to your brief when it is not in the library.

  • Accuracy

    Checked by people

    Metadata built for AI training selections, with people checking a sample before anything is delivered.

  • Speed

    1 to 2 working days

    One working day for buyers, two for content owners.

  • Trust

    Backed on copyright

    Licensed on written terms for AI training. We warrant the rights and indemnify you against copyright claims.

How is video licensed for AI training?

Through a written licence. An AI team sends a brief, checks a sample, then licenses the footage it needs, priced per hour of content, for an agreed use, term and territory. Content owners license their archives through a licensor that represents their footage and pays them a share of each licence.

Bring us a brief, a collection or an archive.

Contact us