Watermarks and authenticity detection examples
Scenario: Detect tampering before adding a sharing watermark
First check this proof image for obvious editing or tampering risk and show me the result. After I confirm, create a copy with a translucent “For review only” watermark.
The agent returns the tamper-detection result first and adds the watermark only after confirmation. The result reflects an algorithmic assessment and confidence level and cannot replace human review or forensic analysis.
Scenario: Check whether an image may be AI-generated
Check whether this image may have been AI-generated and return the decision and confidence information.
The agent returns a structured AI-generation detection result. Treat it as a risk signal rather than the sole proof of image origin.
Scenario: Add a sharing watermark to a PDF
Add a tiled gray DRAFT watermark to this PDF with opacity 0.2, and preserve the original file.
The agent creates a new watermarked PDF using the requested text, color, opacity, and font size.
Scenario: Remove a watermark from an image or PDF
Remove the watermark from this PDF, create a new file, and do not overwrite the original.
The agent selects image or PDF watermark removal based on input type. PDF removal supports up to 100 pages, and results depend on watermark opacity, background, and coverage.