detector.guru 1.5.0: deeper image evidence with Copyleaks

detector.guru 1.5.0 expands Deep Analysis with a new Copyleaks signal and a clearer way to inspect the evidence behind a result.

A stronger second perspective

Deep Analysis now sends the same image to Copyleaks alongside detector.guru's existing detection path. Each provider result remains visible, so you can compare the signals instead of relying only on the combined estimate.

This release also hardens the provider integration: unsupported or incomplete provider responses are handled explicitly, and a failed external check does not get presented as successful evidence.

Localized evidence when available

Copyleaks can return a pixel-level localization mask. When the response contains a valid mask, detector.guru decodes it and overlays the reported regions on the analyzed image. You can switch the overlay on and off while reviewing the result.

Not every image or provider response includes localized evidence. The absence of an overlay does not by itself indicate that an image is human-made.

A clearer result workspace

The result view now keeps the image, provider scores, combined estimate, and available localized evidence together. Deep results remain saved in your private History, including upgrades from an existing Standard analysis.

The upload experience has also been refined: you can upload, drag, or paste supported images and prepare multiple files before choosing Standard or Deep Analysis.

Detection is evidence, not proof

AI-image detection remains probabilistic. Editing, compression, screenshots, unfamiliar generators, and differences between models can affect the output. Use every score and overlay alongside provenance, source context, metadata, and human judgment.

Try Deep Analysis or revisit completed checks in your private History.