AI Disclosure

Draft2Cut is built around automated generation. AI is involved throughout production: writing and analysing a screenplay, deriving characters, environments and shots from it, drawing characters and their turnarounds, composing storyboards, synthesising voices, composing music, animating faces and generating video.

None of it starts by itself: every generation is explicitly requested by a person. There are no agents acting autonomously, and no processes that produce content without someone asking.

Draft2Cut has no engine of its own: it relies on external providers and on engines you can install on your own infrastructure. The choice is made **per craft and per project** — which model is the screenwriter, which the character designer, which the voice — and is always visible on the screen that generates.

No path silently falls back to an engine other than the one chosen: if the selected model is unavailable or unsuited to the operation, the operation refuses and says so. It is a design rule, and it exists so that you never choose one provider and get billed for another.

A current list of available engines, with each one's provider and country, is available by writing to the addresses at the foot of this page: it changes too often for a published document to keep up.

Results are probabilistic. The same prompt, repeated, gives different results; a character's features can shift between two shots; text can contain false statements delivered confidently; video can produce impossible hands, lettering or perspective.

Draft2Cut reduces the drift — with reference packs that hold appearance steady, and by keeping every generated version so the best can be chosen — but does not remove it. Every result should be looked at before it is used, and none of them replaces professional advice.

It depends on the engine chosen, and the answer is not uniform. For most paid providers with a direct API the answer is yes: what is produced is assigned to whoever requested it.

Some providers, however, take a more qualified position: they claim no ownership, but do not warrant that you own the output, pass the underlying models' licences through to you, and place the risk on you. For self-hosted models the answer depends on each model's licence, and in some cases commercial use is not permitted.

Anyone who needs contractual certainty for a specific project should ask what the intended provider's position is **before** producing, not after: the answer is known and will be given.

These are two separate questions. Ownership under the provider's contract says who may use the file; copyright protection says whether that file is a protected work, and many jurisdictions require recognisable human creative input.

The way one works with Draft2Cut — writing, choosing, correcting, iterating, editing — argues in favour of that input, but it is a judgement to make with a lawyer on the specific case. We do not promise it here.

Engines that make a face speak or clone a voice process biometric data. Anyone uploading a real person's likeness or voice must hold a written release, with a defined scope of use and duration: facial-animation providers do not grant those rights, they demand them from whoever uploads.

Producing material that depicts a real person deceptively, or attributes to them statements they never made, is prohibited by the Terms of Service regardless of any consent obtained.

To generate, the material needed to produce that result is sent to the provider chosen for that craft, and to that one only. No provider is permitted to use it to train its models.

Anyone supplying their own API keys sends content under their own contract with the provider. Anyone choosing the local engines — running on their own computer or server — transmits nothing to anyone: it is the only configuration in which material never leaves your perimeter, and it is available for writing, images and voice.

The detail of transfers, legal bases and retention periods is in the Privacy Policy.

Every generation leaves a record: which activity, which model, how much it consumed and what it cost. It serves to attribute costs, diagnose failures and answer a disputed charge. Generated content stays with whoever requested it and is not used for anything else.

On provider contracts and licences: legal@draft2cut.com. On personal data: privacy@draft2cut.com.