Method

Content violence

Scene violence rating (vision model)

What we measure and how

Sampled scene-change keyframes (or fixed-interval keyframes for static content) are batch-classified by Claude vision against a ladder of violence severity levels: `none`, `mild`, `moderate`, `severe`. Reports the highest level observed vs `constraints.content.classification.violence.max_level` from the spec.

Spec fields consulted

These Universal Media Delivery Profile paths resolve this check's thresholds against your spec snapshot at the moment the job was submitted.

• · constraints.content.classification.violence.enabled
• · constraints.content.classification.violence.max_level

The full UMDP schema: /umdp.

Related methods

• Content explicit
• Content prohibited keywords
• Content scene classification
• Content strong language

Method identifier: content_violence