Where was AI used?
The declaration shows which parts of a production used AI, whether that is a single shot or the whole film, and whether it concerns image or sound.
The format has one main task: enable makers to declare which parts of a film used generative AI, to what extent and with what result. This creates an answer that audiences, crews, festivals and clients can understand.
The format starts with the information people need. Technology only exists to make that information useful and transferable.
The declaration shows which parts of a production used AI, whether that is a single shot or the whole film, and whether it concerns image or sound.
The declaration distinguishes between kinds of AI, from a supporting tool to deploying an AI agent. You can read more about that below.
The same declaration can stay with the film and be used by festivals, broadcasters, funds, clients and archives.
They run from light support to work handed over entirely. Visitors see plain language; the corresponding IPTC term is only included for systems and technical reference.
AI helps with research, ideas, planning or analysis, for example, but does not provide generated material for the finished film.
No AI source term needed for the finished materialCaptured image, sound or text was extended, replaced or corrected using generative AI, for example with generative fill.
IPTC: compositeWithTrainedAlgorithmicMediaHuman-captured or human-created material was combined with elements made using generative AI.
IPTC: compositeSyntheticThe image, sound or text was created entirely or almost entirely with a generative AI model.
IPTC: trainedAlgorithmicMediaAn AI agent autonomously performed part or all of a production task, role or department. The declaration records what was taken over and what human oversight remained.
Process information · no IPTC source termMakers do not need technical knowledge. The form only asks for information that is relevant.
Answer whether and how generative AI was used in each stage, from development and production to editing, sound and finishing.
Where AI was actually used, record which crafts, systems and applications were involved.
The answers become a signed declaration, a short credit statement and a reusable data file.
You do not need to know these terms to create a declaration. Behind the scenes, they make the result recognisable, interoperable and potentially more verifiable later.
A stable number keeps one declaration identifiable when it is shared or archived.
An open, machine-readable file prevents the same information from being entered repeatedly.
Existing international terms help systems mean the same thing when describing media provenance.
A later step can attach selected provenance information to a film, image or audio asset with a signature.
Each technical component is explained below with examples, limitations and possible next steps.
What the local number identifies, why it remains stable and what a future public registry could add.
Read the technical explanation →02What data the download contains and how festivals, archives and other systems can reuse it.
Read the technical explanation →03How existing international terms describe how concrete image, sound or video came into being.
Read the technical explanation →04How provenance can be cryptographically bound to a media asset and what the site can already prepare.
Read the technical explanation →05What is guaranteed to keep working, how to add your own fields, and when 1.0 arrives.
Read the versioning policy →