Image & Design

Illuminarty

Understand Illuminarty AI image detection, probability scores, paid localisation features and why detector output needs corroborating evidence.

What Illuminarty claims to detect

Illuminarty offers AI-content detection. Its public page describes image-generation probability, localisation and model-identification features, as well as text-related detection. The important distinction is between an estimated likelihood and proof of how a particular file was created.

We have not measured the service's false-positive or false-negative rates. This review therefore does not endorse an accuracy percentage, rank it against other detectors or treat a result as a forensic conclusion.

Free and paid features

The public plan table lists a free tier for basic classification and paid tiers for additional features such as localisation and API use. It lists Basic at USD 10 per month and Pro at USD 30 per month. Verify current availability and account entitlements before purchasing; the public page is the source of these figures, not a checkout quote.

If your only question is whether to inspect a file more carefully, a basic result may be enough to guide the next step. If a workflow depends on identifying a generated region, verify that feature and the output format with an example before integrating an API.

How to interpret a detector score

Keep the original file and its context. A probability is not a binary record of provenance. A low score does not prove that a picture is authentic, and a high score is not enough to accuse its creator of deception.

Use several forms of evidence: where the image came from, whether the original is available, what edits were disclosed and whether the creator can explain the workflow. Repeatedly resubmitting a screenshot or recompressed copy can change the input; record which version you checked.

A sensible evaluation process

Before relying on a detector in a decision, build a labelled set for your own context. Include known camera photographs, clearly disclosed generated images and edited examples you have permission to use. Track incorrect flags in both directions. That is a proposed evaluation process, not a test we have completed for this article.

For a school, publisher or employer, a detector result should trigger review and a chance to explain the evidence. It should not be the sole basis for a consequential allegation. Avoid uploading confidential material until you understand the service's data handling.

Verdict

Illuminarty can be considered as one signal in an image-review process. It does not replace provenance checks or human judgement. If you create imagery with DreamUp or edit it with Reface, recording and disclosing your process is more useful than trying to obtain a particular detector score.