The plain statement
Every video clip Scribbo produces is artificially generated. It is not a recording of anything that happened, and no part of it was filmed. The sample clips, gallery stills and illustrative imagery on this website are also generated, and are shown to illustrate what the service does.
Where AI is used — and where it is not
| Step | How it is done |
|---|---|
| Straightening the photo, trimming the edges, lifting the paper to white | Ordinary code. Geometry and colour maths — no model involved. |
| Checking the upload is a drawing and not a photograph of a person | An automated classifier. It can refuse an upload; it never alters one. |
| Animating the character | A generative video model. This is the AI step. |
| Prices, plan names, labels, dates and every other piece of text on this site | Typeset by the website. No text anywhere on Scribbo is drawn or written by a model, so a price can never be garbled by one. |
How clips are marked
Every file you download carries its AI origin with it, in three ways:
- Machine-readable provenance. A C2PA Content Credentials manifest is attached to the file, recording that the clip was generated by an AI system and when. Software that understands Content Credentials — including a growing number of social platforms and image tools — can read this without any help from you.
- Embedded file metadata. The MP4 container additionally carries a plain declaration in its metadata fields, so that tools which do not yet read C2PA can still see it.
- A visible label. The studio labels every render “AI-generated” on screen, the download control repeats it, and Free-plan clips carry a visible Scribbo watermark.
Removing or tampering with any of these marks breaches the Acceptable Use Policy.
Why we do this
Because people should be able to tell. A short film of a child's drawing is a lovely thing to send to a grandparent, and it is also a synthetic video. Those two facts sit together comfortably as long as nobody has to guess.
It is also the law in places. Article 50 of the EU Artificial Intelligence Act requires providers of systems that generate synthetic video to mark their output in a machine-readable format and to make the fact of generation detectable. That obligation applies from 2 August 2026. Scribbo launched after that date and therefore has no transition period — the marking described above was built in from the first clip rather than retrofitted.
What the model does and does not do
- It preserves the drawing. The point of Scribbo is that the character keeps its original wobbly outlines, its colouring outside the lines and its paper grain. The model adds movement; it does not tidy, redraw or “improve” a child's work.
- It is not perfect. Generated video is probabilistic. Limbs can drift, details can wander and two runs of the same drawing will not be identical. Failed and broken renders are returned to your allowance automatically.
- It never sees a face. Photographs of people are refused at upload, so no real person is ever put through the model.
- It does not learn from you. Your uploads and clips are not used to train, fine-tune or evaluate any model — see the Privacy Policy.
Human oversight
Automated checks can be wrong. If an upload is refused and you think it should not have been, email support@joelsonbarros.shop and a person will review it. We would rather be too cautious and occasionally wrong than the reverse — but we do fix it.
Model providers
The underlying image and video models are operated by third-party providers under contracts that forbid them from retaining your content for training. We may change provider as the technology moves; the commitments on this page apply regardless of who is behind the generation step.