Disclaimer
Last updated: October 2026
1. Every result is an automated measurement
Everything DiffALL reports — similarity scores, quality bands and scores, voice verdicts and AI estimates — is produced automatically by software, from the files you upload, by the methods described on each page. No person reviews a file or a result. Results are provided for information only and are not professional, technical, legal or any other kind of advice.
2. Comparison scores
A similarity score says how alike two files are by a specific measure (for example SSIM for pictures, timbre and loudness over time for audio, word error rate for text and subtitles). It is not a guarantee of how the files look or sound to a person, and it depends on what was read: where a plan, a chosen start point or a file's length limits the span analysed, the result says so, and it describes that span only. Two files that start at different points, differ in length or come from different recordings can score low even when they share content.
3. Quality checks
A quality check measures one file without a reference to compare it against, so there is no perfect original to anchor the answer. Each check reports a measurement and a band; the overall score and the estimate of how people would rate a photograph are predictions, not facts about the file. A check can decline to answer where it cannot read a file reliably, and a good band is not a certificate that a file is free of problems.
4. Voice comparison
The voice comparison estimates whether two recordings sound like the same speaker. It is not speaker identification and not a forensic voice analysis. It can be wrong, it is less reliable on short, noisy, compressed or phone recordings, and it cannot be relied on against a disguised or imitated voice. “Not conclusive” is a real answer, not a failure.
5. The AI check
The AI check reports what a file declares about its origin and a model's estimate of how likely it is that AI made it. That estimate can be wrong in either direction: real photographs can score high and AI images can score low. A small AI edit inside a real photograph is rarely caught, video is caught less often than still images, and digital art, heavily edited photos and copies compressed by messaging apps are where mistakes are most likely. Content Credentials and AI metadata only show what the file records about itself: they are easily lost, so their absence proves nothing, and a valid signature shows who signed a file, not that what it shows is true. The detector is updated over time, so the same file may get a different result later.
6. No result is evidence
No DiffALL result is a forensic examination, an expert opinion or a certificate, and none may be treated as evidence or proof in any legal, court, police, regulatory, insurance, employment, academic or disciplinary matter. Do not use a result as the only basis for accusing, reporting, sanctioning, grading or dismissing anyone, or for removing anyone's content. Where something real depends on the answer, check other sources and ask a qualified expert.
7. Provided as is
DiffALL is provided “as is” and “as available”, without any warranty of accuracy, completeness or fitness for a particular purpose. DiffALL is not responsible for any decision made, or action taken, on the basis of a result. The full terms, including the limitation of liability, are in the Terms of Service.
8. Links and articles
The blog and help pages explain general concepts and may link to other websites. They are general information, not advice for a particular case, and a link to another site is not an endorsement of it. We are not responsible for the content of other sites.
9. Changes
This disclaimer may change as the tools change. The date at the top shows when it was last updated. Questions: lucynight.studio@gmail.com.