When you need to establish whether a photograph is authentic, "it looks real" is not enough: you need technical evidence you can show and discuss.
Forensic analysis of a digital image examines two things together: what the file tells (technical data, provenance, saving passes) and what the photograph shows (light, optics, materials, micro-texture). Uthentic reads both and delivers a written report, with suspicious areas highlighted and limits stated — not a score to take on trust.
The file's hidden history is read before the image itself.
Camera model, date, lens, saving passes: when a file keeps its history, that history goes into the report. When it has been wiped or rewritten, the report says so instead of ignoring it — missing data is information too.
Content credentials, when present, are verified and reported.
Some files carry a signed statement about how they were produced or edited. Uthentic reads it, records it in the report and compares it with what the image actually shows: a statement and evidence are not the same thing.
The photograph is read the way a photographer would read it in the studio.
Light direction against the shadows, reflections that repeat — or fail to repeat — the environment, depth of field consistent with the declared lens, skin behaviour under backlight. These are craft checks, and they are the ones generated images fail most often.
The zones that betray generation are highlighted on the photo, one by one.
Repeating textures, edges that are too uniform, details that dissolve: every flagged area carries the reason for the flag, so you can verify the reasoning yourself instead of trusting a number.
The report always distinguishes three situations: a generated image, an authentic image edited in post-production, and insufficient evidence. The second is the distinction that matters most to people who work with images: pushed contrast, cleaned skin or a replaced sky are edits to a photograph, not proof of generation, and they are described as such. The third is a choice of honesty: when the signals contradict each other or the file is too degraded, the report suspends judgement instead of forcing a percentage.
The final document is a PDF with the summary, the flagged areas, a verification checklist and a code that lets whoever receives it check its authenticity. It is built for anyone who needs to show the result to a newsroom, a client, a lawyer or a platform. The analysed file, meanwhile, has already been deleted: it is only used during processing and never feeds any model's training.