內容簡介
AI Image & Video Disclosure Labels 外掛可為圖片和影片添加可見的 AI 標示，協助網站擁有者遵循 EU AI 法規的要求，確保媒體內容的透明性與合規性。
【主要功能】
• 獨立開關可控制圖片和影片的標示
• 每個 Gutenberg 圖片區塊可單獨開關
• 每個 Gutenberg 影片區塊可單獨開關，標示位於播放控制下方
• 支援自訂每個標示的文字內容
• 提供可編輯的預設可見文字
• 選擇性添加機器可讀的 Schema.org 數位來源類型聲明
外掛標籤
開發者團隊
② 後台搜尋「AI Image & Video Disclosure Labels」→ 直接安裝（推薦）
原文外掛簡介
AI Image & Video Disclosure Labels adds visible AI disclosure labels to images and videos. At the top of the settings page you can enable image disclosures, video disclosures, or both. Image labels appear on the image; video disclosures are deliberately placed immediately below the video so they do not cover playback controls. By default, labels remain explicitly opt-in per media use and existing content stays untouched. If you enable the optional Media Library auto-label setting, enabled media that you deliberately classify in the Media Library can receive its matching visible disclosure automatically.
Why disclose AI media?
Article 50 of the EU AI Act distinguishes between providers and deployers of AI systems. Providers of certain generative AI systems must add machine-readable marks to generated or manipulated outputs. Separate disclosure duties apply to deployers in specified cases, including deepfakes and certain public-interest text. The relevant rules apply from August 2, 2026. This plugin helps site owners add a visible disclosure and an optional publisher-supplied structured-data declaration to selected images and videos.
Features
Independent global switches for image disclosures and video disclosures, so a site can use either feature on its own or both together.
Per-image switch for Gutenberg Image blocks.
Per-video switch for Gutenberg Video blocks; the visible disclosure is rendered below the video rather than over playback controls.
Separate per-post/per-page switch for the featured image.
Optional custom text for each marked image or video.
Separate editable default visible texts for AI-generated and AI-modified media, plus optional video-specific wording; per-use and featured-image custom text still takes precedence.
Optional machine-readable Schema.org digitalSourceType declarations using standardized IPTC source categories, emitted as ImageObject or VideoObject as appropriate.
Global default and per-use source type for marked images and videos: created using generative AI, edited using generative AI, or enhanced using AI.
Simplified Media Library classification aligned with plain disclosure language: AI-generated, AI-modified, no AI used, or unclassified. More precise legacy technical provenance is retained internally when available.
Media Library list tools: an AI status column, status filter and grouped bulk actions make it possible to classify many images and videos without opening them individually.
Optional automatic visible disclosures for AI-classified Media Library images and videos; this remains off by default.
Optional AI-assisted image analysis through the WordPress 7.0+ AI Client / configured Connectors when available, or directly through OpenAI, Google Gemini, Anthropic Claude, Cloudflare Workers AI, an OpenAI-compatible HTTPS endpoint, or a custom HTTPS analysis endpoint. Automated analysis remains image-only in the 3.0.x series; videos are classified manually.
Dynamic model discovery where the provider exposes a Models API or, in WordPress AI Client mode, through the configured AI provider registry; model names are never hard-coded and manual model/policy names remain supported.
Provider-independent custom HTTPS endpoint contract for Cloudflare Workers, AI Gateway, dedicated detectors and other future services.
AI analysis is stored separately from the publisher-facing status. Suggestions can be reviewed, filtered and bulk-applied; “likely non-AI” is never converted to “No AI used” automatically.
Optional visual auto-classification uses an administrator-adjustable confidence threshold (95% recommended default). The confidence value is the model’s self-assessment, not a calibrated forensic probability.
A separate custom-endpoint option can accept technically verified provenance only when the administrator’s own endpoint actually verifies a provenance/watermark signal; ordinary model confidence never counts as verified provenance.
Background bulk analysis for selected images, all unclassified images or the entire Media Library, with image-count and known-cost safety limits.
No hard-coded API pricing. Cost information is taken from provider/custom-endpoint data when machine-readable, calculated from user-supplied token rates, inferred from previously observed request costs, or shown as unavailable.
Four image-overlay positions and three design presets, plus detailed design controls with live image preview.
Optional separate video badge design with its own alignment, colors, opacity, border, radius, typography and padding, plus a dedicated below-video preview. If disabled, videos inherit the image badge design.
Responsive behavior based on the actually rendered image width: hide the label on tiny thumbnails, show a compact AI symbol on medium images, and the full text label on large images. An optional touch-first mode prefers the compact symbol on phones and tablets by using browser interaction capabilities rather than user-agent guessing.
Location-specific display rules: force icon-only mode or hide an existing disclosure inside selected page areas by using CSS selectors.
Featured-image support across homepages, archives, query blocks, cards and post loops, including changing or randomly selected posts.
Three built-in SVG symbols (AI Monogram, Sparkle, Chip) plus a custom PNG/SVG symbol from the Media Library.
Symbol size in pixels or as a percentage of the rendered image width.
Optional disclosure popover for compact symbols: hover or keyboard focus on desktop, tap-to-toggle on touch devices.
Theme integration: works with the WordPress default themes and common featured-image markup out of the box; unusual themes can be supported by adding CSS selectors in the settings — no code required. Developers can additionally use the gdaiidl_featured_selectors and gdaiidl_post_types filters.
Compatible with common page caches: labels are rendered server-side where possible, and supported page caches are cleared automatically when settings or labels change.
Clean uninstall: removes all options and post metadata when the plugin is deleted.
Important legal note
This plugin can provide both a visible disclosure label and an optional publisher-supplied Schema.org declaration in the page source. The structured data uses standardized IPTC Digital Source Type categories, but it does not modify the media file, create or verify C2PA Content Credentials, prove origin or authenticity, or replace machine-readable markings supplied by the provider of a generative AI system. Whether and how specific content must be labeled depends on your role, your content and the applicable law. This plugin is a tool, not legal advice.
External services and privacy
The core disclosure/label features work without any external AI service. AI-assisted image analysis is optional and disabled by default.
When an administrator enables AI analysis and requests an analysis (or explicitly enables automatic analysis of new uploads), the plugin sends a temporary resized copy of the selected image and a short classification prompt from the WordPress server to the configured provider. The original Media Library file is not modified. The plugin stores the returned suggestion and related metadata such as confidence, provider/model identifiers, explanation, usage/cost information and analysis time in WordPress attachment metadata. Images may contain personal, confidential or copyrighted material; administrators are responsible for ensuring they are allowed to send that material to the chosen service.
WordPress AI Client / Connectors (WordPress 7.0+): When this mode is selected, this plugin does not store or read a duplicate provider API key. WordPress routes the request through an AI provider plugin configured under Settings > Connectors. The external service that receives the image is therefore the provider selected by WordPress; review the terms and privacy policy of every connector/provider you enable for AI use.
Built-in service adapters:
OpenAI API – service: https://platform.openai.com/ ; policies/terms/privacy: https://openai.com/policies/
Google Gemini API – service/docs: https://ai.google.dev/ ; API terms: https://ai.google.dev/gemini-api/terms ; Google Privacy Policy: https://policies.google.com/privacy
Anthropic Claude API – service/docs: https://platform.claude.com/ ; privacy/commercial policy center: https://privacy.anthropic.com/
Cloudflare Workers AI – service/docs: https://developers.cloudflare.com/workers-ai/ ; Cloudflare policies: https://www.cloudflare.com/policies/
The plugin can also connect to an administrator-supplied OpenAI-compatible HTTPS endpoint or custom HTTPS analysis endpoint. These URLs must use HTTPS. The plugin cannot know the operator’s terms or privacy policy, so the site administrator must review and disclose the policies of that endpoint.
Direct-provider API credentials configured in this plugin are used only on the WordPress server and are not exposed to site visitors or localized to browser JavaScript. In WordPress AI Client mode, provider credentials remain managed by WordPress Connectors and are not copied into this plugin. The plugin itself does not add telemetry or send images anywhere merely by being installed or activated.
Support
For support, visit drissner.media/kontakt. If the plugin is useful to you, you can support its development through PayPal.
