How we build with AI, responsibly.

How AI content is generated

VidMints is a web-based studio that orchestrates a range of specialised AI models rather than a single one. When you run a generation, your prompt or uploaded media is sent to the model best suited to the task, the result is processed in our cloud rendering pipeline, and the finished file is returned to your library. At a high level we build on:

• Video & image generation — text-to-video, image-to-video, scenes, motion, face and style swaps run on hosted diffusion and video models (via providers such as Replicate).
• Voice & audio — AI voiceover and text-to-speech use neural voice models (such as ElevenLabs) with a fixed catalogue of licensed voices.
• Speech & language — transcription and captions use speech-to-text models (Whisper-class), and scripting/assistant features use large language models.

Model providers can change as better options become available. We choose models for quality, safety and reliability, and we do not represent AI-generated output as a real recording of any real event or person.

Labelling & watermarking

Content made on VidMints is AI-generated, and we think it should be clear where it came from. Free exports carry a small "VidMints" watermark that identifies the origin of the file. All paid plans export clean, watermark-free media for professional and client work — but removing the watermark does not remove your responsibility to use the content honestly and to disclose AI generation where the context calls for it (for example, in news, advertising, or anything that could mislead a viewer).

Safety filters and human review

We combine automated safety filters at the generation layer with human review when content is reported:

• Prompts and uploads are screened for clear violations (such as sexual content involving minors or instructions for serious harm) and blocked before any output is produced.
• Anyone can report an output or account; reports are reviewed by our team and can lead to content removal, restrictions, or termination.
• Automated filters are imperfect. If you believe a generation was wrongly blocked, or wrongly allowed, you can appeal or report it — see our Content Moderation Policy.

No deepfake abuse

Face swap, talking-avatar and lip-sync tools are powerful, and they are easy to misuse. VidMints strictly prohibits using them to create deepfakes of real people — celebrities, public figures, or private individuals — without their explicit consent, and always prohibits sexual, defamatory, harassing, deceptive, or election-related impersonation. Non-consensual intimate imagery and any sexual content involving minors are never permitted under any circumstances. Accounts used for this kind of abuse are terminated, and serious cases (such as CSAM) are reported to the relevant authorities. The full list of prohibited uses is in our Acceptable Use Policy.

How your data is handled

The media you upload is used to produce the output you asked for, and nothing more. Specifically:

• We do not use your uploads or generations to train general-purpose AI models.
• We do not sell your data or share your files with advertisers.
• Files are processed by trusted infrastructure and model providers solely to run the compute you triggered, and stored in your private library until you delete them.

For the full detail on what we collect and your rights over it, see the Privacy Policy, and for how we protect it, see the Security overview.

You own what you make

You keep ownership of the media you upload and, subject to our Terms, of the AI outputs you generate. Plans that include a commercial usage license let you use those outputs in monetised and client work. In return, you are responsible for making sure your inputs and outputs respect other people's rights and comply with the law.

Questions about our AI practices, or something to report?

Email trust@vidmints.com or use the Contact page. Related reading: Acceptable Use · Content Moderation · Security.