A video takes minutes to render.
The decision takes 100 ms.
Text-to-video and image-to-video prompts are checked against your policy before they reach the render queue, along with the caption of any reference image. Adult content on your terms, real people held for a human, minors never. Nothing has to be rendered to find out what it was.
Starts from the adult-platform preset.
Video carries the same lines as stills, with more at stake per render. The preset tags adult work instead of blocking it and sends named or described real people to review before a frame exists. If likeness is the line your platform cannot cross, move real people from review to block: it's one line in the policy.
- ✓ Every line is yours to move, except the minors line
- ✓ Every decision explains which checks fired and why
- ✓ Dry-run any policy change on live traffic before you enforce it
The risk: likeness in motion.
A clip of a real person saying or doing something they never did is the deepfake everyone worries about, and a video that exists can be screen-recorded and reposted long after you delete it. Frame-by-frame moderation after the render is slow, it costs the GPU time you already spent, and it is too late. Vindex puts named and described real people in front of your reviewers before the job starts, and blocks anything sexual involving a minor outright, in every language.
Vindex for other platforms
See your own prompts scored, live, in twenty minutes.
Bring a hundred prompts you've argued about internally. Start the trial and run them yourself to see where your policy line actually sits.