AI Content Policy
IntactShot turns a product photograph — or, for software advertisers, an app or interface screenshot — into a short video, in a frame shape you choose, in which a presenter talks about it. Because the presenter is generated, this page sets out exactly what the system can and cannot produce, where its material comes from, and what we do not permit. It is operated by Fatih Sancar.
1. The presenters are synthetic and depict no real person
Every avatar in the library was generated by us from a written text description, using a text-to-image model. None is a photograph of a real person, none was modelled on a real person, and none is intended to resemble any real, living, deceased or public individual. Each prompt states this explicitly as a generation constraint.
The library is fixed and closed: it contains a set number of presenters that we created and host ourselves. Users choose from that library. They cannot add to it, alter a presenter’s face, or supply their own.
2. What the Service cannot do
These are absent by design, not disabled by policy:
- No face upload. There is no face-upload feature. A user can supply two kinds of image: the thing being advertised — a product photograph or an app screenshot — which is only ever used as the advertised object, and optionally a photo of an empty place to film in, which is only ever used as the setting. Every setting photo is checked before it is saved and refused if a person or face is detected in it. Neither ever becomes the presenter. Nothing a user uploads is animated, lip-synced, or used to shape a face, so a photograph of a person supplied as a “product” still cannot be made to speak.
- No face swap. No feature substitutes one person’s face for another.
- No voice cloning. There is no route to upload a voice sample. With the default renderer, speech comes from a fixed catalogue of preset voices licensed to us (see section 3); a user selects one and cannot create, upload or modify a voice. A user may instead choose a video model that generates its own sound; that model is given the script and the frame, never a voice sample or a named person to imitate.
- No deepfakes or impersonation. The Service cannot be directed to reproduce the appearance, voice or identity of a named or depicted individual. There is no prompt path, setting or API by which a user can request one.
3. Where the material comes from
- Presenter images: generated by us with a commercial text-to-image model under our own account, from prompts we wrote, and stored on our own infrastructure. We hold the provenance record for every one.
- Voices: preset voices licensed to us through a commercial text-to-speech provider’s voice marketplace. Under that provider’s voice library terms, a creator publishing a voice must certify that it is their own voice and not another person’s, that they hold the rights to license it, and that they consent to its use for speech synthesis by the provider’s customers; the provider reviews voices before listing them, and creators set their own licensing terms and may withdraw a voice. We select only from the provider’s reviewed voice tiers and deliberately exclude its cloned-voice and celebrity/public-figure categories. We do not clone, alter or create voices ourselves, and the Service gives users no way to do so.
- Product images, app screenshots and setting photos: supplied by the user, who warrants that they own them or are authorised to use them.
- Models: we do not train models. We use commercially available third-party models through their providers’ APIs and do not fine-tune them on user data or on any dataset of real people.
4. How the product is represented
Worth stating because it is unusual, and worth stating precisely. A generative video model reinterprets whatever it draws, so a product or an interface it invents from a written description drifts as the clip runs. We therefore never ask a model to imagine the goods. The image the user uploaded is separated from its background and given to the image model as the visual reference for the still frames the video is animated from, rather than a description of it; the video model then animates those frames, so the product can drift as the video runs.
For software advertisers the same applies to the screen. The uploaded screenshot is the reference for the device screen in those frames, and that screen is model output too: small text on it can drift from the original. A user who needs an interface to read verbatim should check the delivered video before running it.
For a physical product, every frame the viewer sees is generated. The presenter, the setting and the item in their hand are all model output, produced with the user’s own photograph as the reference. It is a depiction of the goods, not a photograph of them, and a viewer should read it the way they read any advertisement: as a representation. Users are responsible for the accuracy of what they advertise (section 5), and a user who needs a guaranteed-exact rendering of a label should photograph it rather than generate it.
This section changed on 7 September 2026. It previously stated that a physical product’s cutout was composited over the finished video as a card of its own original pixels. That overlay was removed from the product, so the claim was withdrawn rather than left standing.
5. What users may not do
Using the Service for any of the following is a breach of these terms:
- advertising a product the user does not sell, or does not have the right to advertise;
- suggesting that a presenter is a real customer, a real employee, a real expert, or a real person of any kind — presenters are synthetic and must not be presented as genuine testimony;
- implying that any real person or organisation endorses a product when they do not;
- making claims that are false or misleading, including health, medical, financial or performance claims;
- promoting goods that are illegal, counterfeit, stolen or regulated;
- producing sexual, hateful, harassing, violent or otherwise unlawful material;
- generating content that features or targets a minor;
- removing, obscuring or defeating any disclosure that content is AI-generated where the law or the destination platform requires one.
6. Disclosure
Videos made with IntactShot are AI-generated and the presenter is not a real person. Several jurisdictions and most social platforms now require that synthetic media in advertising be labelled. Meeting that requirement on the platform where the video is published is the user’s responsibility, and we recommend disclosing in every case, including where it is not yet mandatory.
7. Enforcement
Generation requests are subject to the safety filters of the underlying model providers, which reject prompts and images that violate their policies. On top of that we investigate every report we receive, and we may review the content of an account where we have reason to believe this policy has been broken.
Where we find a breach we may remove the content, suspend the feature, or terminate the account without a refund, in proportion to what happened. Where content appears to involve someone’s likeness or a criminal offence we will act on it regardless of the account’s standing.
8. Reporting a problem
If you believe a video made with IntactShot misuses your likeness, infringes your rights, or breaks this policy, email support@intactshot.com. Tell us where you saw the content and what the problem is. We acknowledge reports within 72 hours and we prioritise anything involving a real person’s likeness or safety.
9. Changes
As the underlying models and the law around synthetic media change, this policy will change with them. The version in force is always the one on this page, with the date it took effect at the top.