Ship it, disclosed. The six checks before anything renders.
The rights and disclosure checklist that runs before anything ships: likeness, music, source material, platform AI labels, client scope.
TikTok and YouTube require creators to label realistic AI-generated video, and Meta applies automated AI Info labels. Before a video with an AI character ships, we put an on-screen AI-character card in the opening seconds, say so in the caption, and switch on the platform's AI label wherever one exists.
Pre-flight disclosure check
Six checks from the list below. Tick what your render passes.
Disclosure is not paperwork. It is the part that gets a finished roll pulled, demonetized, or rejected by a client after the fact. Below: the checklist we run before anything ships, written from a pipeline that documents every run with a recorded ledger at makeroll.com/#pilot.
Why a checklist at all
Big brands ship AI video at scale and publish the receipts. Google's own case study documents a Qatar Airways AI film with 35M+ views (Google case study). WPP documents Cadbury's Shah Rukh Khan My-Ad campaign at 130,000 personalized ads and 94M views (WPP case study). Those campaigns cleared disclosure and rights because the work happened before the render. Yours should too.
Each one has killed a real render
Better here than after publish.
- Likeness. If a rendered person resembles a real, identifiable one, you need consent or you need a different face. Never imply endorsement by Qatar Airways, Cadbury, or any featured creator. For synthetic characters, use the exact pattern below.
- Music. Licensed or royalty-free, with the license saved next to the project file. A video that decodes cleanly but rides on an uncleared track does not ship.
- Source material. Every image or clip you feed a model has its own rights. If you did not make it, check what license lets a model ingest it and lets you publish the output.
- Synthetic-media labels. Turn on the platform's AI-content tag when the output is synthetic. Label it even when the character is fictional, especially when the character looks real.
- Client use scope. A clip cleared for a personal feed is not automatically cleared for paid media. Say what the license covers, in writing, per deliverable.
- Platform AI labels. Each platform's tagging rules differ and they change. Check the current policy at upload time, not at render time.
What compliant looks like
Keep it short and put it where viewers see it. The exact pattern our own campaign runs on every published demo.
When a fictional character could be mistaken for a real person, the label is the whole defense.
The actual platform rules, by name
The label requirements are published, not folklore, and they changed recently. We checked all of these on 22 Sep 2026.
| Platform | What is required | Where the label sits |
|---|---|---|
| TikTok | Creators must label realistic AI-generated content. Automatic labels apply to uploads carrying C2PA Content Credentials. Since 21 Jul 2026, TikTok Ads requires a disclosure label on every ad featuring realistic AI-generated content. | Upload-time label; C2PA auto-labeling |
| YouTube | Creators must disclose realistic altered or synthetic media, including AI-generated faces and voices, using the altered-content tool in Creator Studio. The label sits in the expanded description, and on sensitive topics like news, health, elections, and finance it shows right on the player. | |
| Meta | Automated "AI Info" labels apply to photorealistic AI video and realistic synthetic audio on Instagram and Facebook, per Meta's AI labeling policy. |
Beyond the platforms, US FTC endorsement guides treat an undisclosed material connection as deception. The EU AI Act's transparency obligations, Article 50, applying from 2 Aug 2026, require machine-readable marking of synthetic media.
Disclosure is a pre-render step
Run the rights check while you still have options. The weekly club pairs one brief with a rights check on every roll (what you get, the pilot, the receipts). Membership is open now (see the plans).
Every figure traces to a named source
Every figure on this page traces to a named source, and the pilot numbers come from a recorded metered ledger, not estimates. The ledger shows 47 metered API calls producing 3 concepts and 6 exports with QA scores of 91 to 96 on 21 Sep 2026 (pilot ledger). Its spend figure covers one recorded pilot's metered research and generation only. It excludes Codex access, human time, subscriptions, media spend, taxes, and retries.
Reel play counts are dated snapshots checked 22 Sep 2026: @bennettwaisbren at 1,019,816,251, @cat_vlog365 at 495,021,428, and @stat.us.ai at 213,154,735 plays. Riffkit's 2026 dataset reports the median AI TikTok receives 418 views, as cited on the makeroll.com landing page. We cite it as the honest baseline because it is. No view or revenue promises here.
What Makeroll actually discloses
Our own UGC campaign runs an original fictional AI character and discloses the same way every time:
- An on-screen "AI CHARACTER · ORIGINAL MAKEROLL DEMO" card within the opening seconds.
- A caption stating the character is an original fictional adult AI character created for the demo.
- The platform AI label enabled wherever the platform offers one.
That is the bar this checklist enforces. If a video cannot pass those three checks, it does not ship.
The camera is optional. Taste is not.
One brief, one finished roll, critique each week. Sources checked 22 Sep 2026. Methodology is public on the pilot ledger.