Screen the prompt, not the image.
Before a single pixel renders.
Image platforms live and die on one question: what are people making? Vindex reads every prompt, negative prompt and reference-image caption against your policy in about a hundred milliseconds, so your GPUs never spend a second on something you would have to take down.
Starts from the adult-platform preset.
Built with an image platform that allows adult content and can never render a minor. Explicit adult work is tagged, not blocked. Real people go to your reviewers. Copyrighted characters are tagged and allowed. If your generator is safe-for-work only, drop the adult line to level one and keep everything else.
- ✓ 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: an image you can't take back.
A generated image exists the moment it renders: in your storage, on your CDN, in a user's camera roll. A keyword list flags "girl" and "25 years old" and still misses the prompt that spells an age out in slang or another language. The only safe place to stop the image that matters is before it exists, and the only defensible way to allow everything else is a written policy you can show a payment processor, an app store or your own team.
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.