On-device intelligence

AI transparency

ImageSage uses on-device machine learning for recognized text, visual search, and optional filename suggestions. It does not send your photos or prompts to an ImageSage cloud service.

Live generative AI

Magic Rename uses a locally packaged generative AI model to create a short proposed filename from a photo you select. The proposal is advisory. ImageSage displays it in an editable review dialog and does not rename a file until you explicitly include and confirm the proposal.

Other machine-learning features

OCR combines local text detection and recognition to make visible words searchable. Visual search uses local image and text embeddings to estimate semantic similarity. These features rank or retrieve your own local images; they do not publish content or make decisions about people.

Limitations

Machine-learning output can be incomplete, inaccurate, biased, misleading, or inappropriate. OCR may misread text, visual search may return weak matches, and Magic Rename may describe a photo incorrectly. Do not rely on ImageSage output for safety-critical, legal, medical, financial, identity, or accessibility decisions.

Report inappropriate output

Send only the generated words, ImageSage version, and a general non-identifying image category. Do not send the source image, private text, names, or paths.

Compose report

Windows opens your default mail app with a pre-addressed draft. Nothing is sent unless you choose Send.