Responsible use

Kalki generates synthetic video. These rules are part of its design, and some are enforced by tests/policy.test.ts.

Provenance stays on

The Producer tags every file with comment=AI-generated content (created with Kalki) and encoder=Kalki. Kalki does not remove C2PA manifests, SynthID or any other provenance signal a provider embeds, and it does not write fake camera, device or date metadata. Contributions that do any of these will not be merged.

Disclose when you publish

Uploading is manual. When you post a Kalki video, turn on the platform's disclosure setting for altered or synthetic content (YouTube has one), and follow its policies on AI-generated media, spam and repetitive content.

Providers on their own terms

Kalki calls each provider only with an API key you supply, under your account. It has no code to create accounts, solve CAPTCHAs, rotate proxies or evade bot detection.

People and likeness

Characters are fictional by default. Do not create characters that impersonate real people, and do not use reference images of anyone without their consent.

Reporting misuse

If you find a feature that makes misuse easier, open an issue. For a security problem, follow the security policy.