TencentCloud Image AIGC Detection

Problem

Online content platforms, e-commerce review, and UGC communities often need to determine whether an image is authentic or produced by a generative model. Manual sampling is expensive, and fixed rules struggle to cover the output styles of Stable Diffusion, Midjourney, and GPT-4o, which can let synthetic images slip through before publication. This skill wraps Tencent Cloud image content security AIGC detection into an executable capability for image authenticity screening, AI artwork review, and synthetic media checks.

How It Works

• It accepts an image URL or local file path; local files are read and converted to Base64.
• It calls the ImageModeration API with the IMAGE_AIGC detection type instead of relying on an LLM to guess.
• It returns suggestion, label, score, and detail_results: Pass indicates an authentic image tendency, Review indicates manual rechecking, and Block indicates likely AI generation.
• Before running, configure Tencent Cloud API credentials and set TENCENTCLOUD_AIGC_RECOG_IMAGE_BIZ_TYPE; add TENCENTCLOUD_TOKEN when using temporary credentials.

Boundaries

Supported formats include PNG, JPG, JPEG, BMP, GIF, and WEBP, with a 5MB image limit. API rate depends on console quotas. Results depend on the account policy ID, and the agent should not fabricate outcomes when the API fails; it should return the error to the caller.