AI Providers
Updated
Pro feature. Available in MediaVerse Pro.
Requires WPMediaVerse Pro - This feature is available exclusively in the Pro version.
WPMediaVerse Pro adds Google Cloud Vision, AWS Rekognition, and Claude (Anthropic) as AI analysis providers alongside the built-in OpenAI Vision option. All four providers support auto-tagging and content moderation.
Supported Providers
| Provider | Auto-Tag | Content Moderation | Configuration Location |
|---|---|---|---|
| OpenAI Vision (free + Pro) | Yes | Yes | Media > Settings > AI & Moderation |
| Google Cloud Vision | Yes | Yes | Media > Settings > AI & Moderation > Google Vision |
| AWS Rekognition | Yes | Yes | Media > Settings > AI & Moderation > AWS Rekognition |
| Claude (Anthropic) | Yes | Yes | Media > Settings > AI & Moderation > Claude |
Only one provider is active at a time. Select it under Media > Settings > AI & Moderation > Provider.
Google Cloud Vision
Requirements
- A Google Cloud project with the Cloud Vision API enabled
- An API key or service account with
roles/cloudvision.user
Settings
| Option | Option Key | Description |
|---|---|---|
| Vision API Key | mvs_pro_google_vision_key |
Google Cloud API key |
What Google Vision Returns
- Labels: Object and scene descriptions, mapped to
mvs_tagterms when Auto-Apply Tags is on - Safe Search: Likelihood scores for adult, violence, racy, medical, and spoof categories
- Web Detection: Public web entities that match the image (used as supplemental tags)
AWS Rekognition
Requirements
- AWS account with the Rekognition service enabled in your chosen region
- IAM user or role with
rekognition:DetectLabelsandrekognition:DetectModerationLabelspermissions
Settings
| Option | Option Key | Description |
|---|---|---|
| Access Key ID | mvs_pro_aws_access_key |
AWS IAM access key ID |
| Secret Access Key | mvs_pro_aws_secret_key |
AWS IAM secret access key |
| Region | mvs_pro_aws_region |
AWS region, e.g. us-east-1 |
What Rekognition Returns
- Labels: Hierarchical object and scene labels with confidence scores, mapped to
mvs_tagterms - Moderation Labels: Content categories (nudity, violence, drugs, etc.) with confidence scores used by the moderation threshold settings
Claude (Anthropic) (New in 1.8.0)
Requirements
- An Anthropic account and API key with billing enabled
Settings
| Option | Option Key | Description |
|---|---|---|
| Anthropic API Key | mvs_pro_anthropic_key |
Your Anthropic API key |
| Claude Model | mvs_pro_anthropic_model |
Vision-capable model used for analysis, tagging, and moderation. Choices: Claude Haiku 4.5 (fast & low cost, recommended, default), Claude Sonnet 4.6 (stronger judgment), Claude Opus 4.8 (most capable) |
What Claude Returns
- Description + tags: same natural-language description and comma-separated tag suggestions as the OpenAI path, mapped to
mvs_tagterms when Auto-Apply Tags is on - Moderation: flags content against the site's configured AI Flag Criteria and Custom Flag Terms - the same category checkboxes and custom-terms field the free plugin's OpenAI path uses
Circuit Breaker
WPMediaVerse Pro implements a circuit breaker for all AI providers. If a provider records 5 consecutive failures (network timeout, invalid API key, rate limit), the circuit opens and no further API calls are made to that provider for a cooldown period of 1 hour. This prevents quota exhaustion during outages.
While the circuit is open, calls to that provider are skipped. The failure counter is stored in a transient that expires after the cooldown, and a successful call resets the counter immediately. There is no separate half-open/retry phase or manual reset command.
Auto-Tagging Behaviour
All providers map their returned labels to mvs_tag taxonomy terms. Terms are created if they do not exist. Auto-tagging and the confidence/threshold controls that govern it are configured through the free plugin's AI settings (Auto-Apply Tags and the moderation thresholds under Media > Settings > AI & Moderation) - the Pro providers feed into that same pipeline rather than adding their own tag-confidence option keys.
Content Moderation Thresholds
Each safety category has an independent threshold. When a category's confidence score meets or exceeds the threshold, the moderation action configured in When AI Flags Content is applied.
Configure thresholds at Media > Settings > AI & Moderation > Moderation Thresholds.