AI Moderation

Free + Pro - Core functionality is included free. Features marked with (Pro) require MediaVerse Pro.

MediaVerse integrates with OpenAI Vision (GPT-4o Mini by default) to automatically analyze and moderate uploaded media. MediaVerse Pro adds support for Google Vision, AWS Rekognition, and Claude (Anthropic).

How AI Analysis Works

When Auto-Analyze Uploads (MediaVerse > Settings > AI, off by default) is enabled, MediaVerse sends each uploaded image to OpenAI immediately after storage. The analysis returns:

• A natural-language description of the image
• Suggested tags (comma-separated keywords)
• Safety categories with confidence scores (adult content, violence, etc.)

The description is saved to the ai_description media metadata key. If Auto-Apply Tags is enabled, the suggested tags are assigned to the mvs_tag taxonomy.

How AI Moderation Works

When AI Moderation is enabled (MediaVerse > Settings > Moderation, off by default), MediaVerse checks each new upload with AI. If it flags the upload, the action chosen in When AI Flags Content is applied. The default is Hide until I review it.

You choose what the AI looks for with AI Flag Criteria (nudity or sexual content, violence or gore, hate or harassment, self-harm, drugs, spam; all on by default, and at least one always stays on). Custom Flag Terms lets you add your own comma-separated terms, such as weapons or gambling.

Action	What Happens
Hide until I review it	Media is hidden from everyone except its author and moderators, and appears in the moderation queue. Approving it restores it exactly as the member left it.
Reject (move to draft)	Media is moved to draft
Delete permanently	Media and its files are removed from local and cloud storage. Cannot be undone.

Before 2.6.0 there was a separate Hide choice that also forced the item’s privacy to private, and approving it did not change it back. Sites still set to it now behave like Hide until I review it; the mvs_moderation_hide_sets_private filter restores the old behaviour.

Moderation status changes fire the mvs_moderation_changed action hook ($media_id, $status, $old_status, $user_id).

Triggering Analysis Manually

Moderators can re-analyze any media item via the REST API:

curl -X POST https://yoursite.com/wp-json/mvs/v1/moderation/123/analyze \
  -H "X-WP-Nonce: NONCE"

Approving or Rejecting from the Moderation Queue

1. Go to MediaVerse > Moderation. The AI Flagged tab lists what AI flagged, Pending Review lists items waiting for a decision, and Resolved / Rejected lists the rest.
2. Review the flagged item and its AI analysis result.
3. Click Approve to publish or Reject to move to draft.

Budget Control

Set a monthly spending cap at MediaVerse > Settings > AI > Monthly AI Budget ($). The default is 10. MediaVerse counts an estimated $0.01 per call and stops making AI calls when the estimate reaches the budget; they resume next month. Set it to 0 for no limit. Real cost depends on the model, so also set a billing limit with your AI provider.

To check current spending:

wp mvs stats

Registering a Custom AI Provider (Developer)

Implement the AIProviderInterface and register your provider:

use WPMediaVerse\Services\AIProviderInterface;

class MyCustomProvider implements AIProviderInterface {

    public function get_id(): string {
        return 'my_provider';
    }

    public function is_available(): bool {
        return (bool) get_option( 'my_provider_api_key' );
    }

    public function analyze( int $media_id ): array|WP_Error {
        // Return array with 'description', 'tags', 'flags'.
    }
}

add_action( 'mvs_ai_providers', function( $ai_service ) {
    $ai_service->register_provider( new MyCustomProvider() );
} );

Moderation & AI Metadata

These keys live in MediaVerse’s mvs_media_meta table (accessed via the MediaRepository), not in WordPress post meta. moderation_status is also a column on the mvs_media_index table.

Meta Key	Values	Description
moderation_status	approved, pending, flagged, rejected	Current moderation status
ai_description	string	AI-generated description
ai_tags	string[]	AI-suggested tags
ai_confidence	float	Confidence score for the AI analysis