AI disclosure
The parts that need saying, said on the page
What gets AI video into trouble is rarely the technology. It is what nobody bothered to say out loud. This page covers our defaults, the limits of what this thing is, and what we do and do not touch.
Masters carry an AI-generated badge by default
The badge is genuinely burned into the frame, not a boolean in the metadata. Every master gets a semi-transparent one in the lower-right corner. You can change the wording to your own brand; turning it off is your compliance decision, not a default we made on your behalf.
Where the limits are
We do the orchestration. We do not make the pixels
Footage comes from external models
xAI, Google, ByteDance, Alibaba, MiniMax. Quality, limits and moderation policy are theirs. Our work is arranging them into a line that actually delivers, and moving to another one when a vendor stops co-operating.
Draft footage is a concept board
Structure, pacing, voice and subtitles are real; the picture is not finished imagery. Its job is to let you judge whether the structure and rhythm are right. Only the final grade goes to real generative models, and the interface says so too, not just this page.
Direct vendor connections only
Aggregators mark up the vendor's own price, so we do not buy through them. Someone else's packaged product is already aggregating, so reselling it means paying two margins. And a grey route that turns consumer subscriptions into an API can be cut off unilaterally on any given morning, which is not a supply chain.
We do not train our own models
All footage capability runs through external vendors. We do not train models and we do not claim to. Adding a vendor is two methods, one catalogue entry and a set of contract tests, which is exactly why swapping one out is routine rather than an event.
Division of labour
Which parts of this line are ours
Saying this plainly is more use than calling ourselves an AI video platform. Where the capability ends is what tells you when to come to us and when not to.
What we do
- Orchestrate an idea into eleven stages, each re-runnable on its own
- Route per shot and fall back per shot — one vendor fails, the next takes it
- The timing arithmetic: voice re-timing shots, dissolves overlapping rather than adding
- Timeline assembly, transitions, burned-in subtitles and rendering the master
- Account for every external call and quote each step before it runs
- Hard tenant isolation — someone else's resource comes back as “does not exist”
What we do not do
- Train our own generative models
- Make the pixels; external vendors do that
- A general video editor: no layers, keyframes or colour curves
- Buy through aggregators, or resell someone else's packaged product
- Touch your card number; payment runs on Stripe's hosted page
- Decide for you whether the AI badge comes off
Your material
Defaults set to the most conservative setting
Masters are private by default
Sharing is something you switch on. We will not introduce a “generated means public” default.
Tenants are hard-isolated
Another tenant's resource comes back as “does not exist”, never as “not allowed”. The second answer confirms the id exists, which is a channel for enumerating other people's work.
We never touch card numbers
Payment runs on Stripe's hosted page and subscription state is driven by signature-verified webhooks. Card details never reach our backend.
Every external call is accounted for
Which vendor, how many seconds or stills, estimated cost. All recorded. You can always find out what a given film cost to make.
Have a specific compliance question?
Commercial scope, data handling, disclosure rules. If your use needs something more definite in writing, open an account, run one episode, then come back with the specific question. That gets further than any page of general assurances.