A small clip can carry a caption that says the opposite of what the person in the video actually says. FactLens is a working system: paste an Instagram link and it listens to the video, reads the caption and the on-screen overlay text, and judges each claim against the content itself. It runs on local AI models: an open-source downloader, a local speech model for transcription and translation, local OCR for the overlay text, and a local language model for the judgment.
Case one, flagged: a public Instagram reel by Md shafi showing a Ministry of External Affairs press briefing, with an overlay presenting a strong official statement as the spokesperson's own words. The speaker in the video only says he has no information to share. The system judged both the caption and the overlay UNRELATED to what is actually said: a quote the speaker never said, presented as his words. Analysed twice, 25-Jul and 26-Jul-2026, with identical verdicts; likes rose from 102,197 to 103,478 overnight, with 45k shares and 13.5k reposts shown on the post.
Case two, cleared: a public Instagram reel by Akash Rajawat (5,735 likes at the time of analysis) fact-checking two viral Sonam Wangchuk clips. The system transcribed the Hindi speech, translated it, read the overlay text, and judged both the caption and the overlay SUPPORTED: that reel is honest. The check works in both directions. Each run took two to three minutes on a desktop computer at near-zero cost.
Only the platform has the video, the caption, the overlay and the audience graph in one place, at marginal cost near zero, before a clip goes viral, with no scraping walls. Independent fact-checkers cannot do this at scale: Meta ended its third-party fact-checking program in the US on 7 January 2025 and moved to Community Notes; Meta's own Oversight Board warned in March 2026 that Community Notes are not a proper global substitute; and the IFCN's State of the Fact-Checkers 2025 report found 45.3% of fact-checking organisations reporting revenue declines, with only 22.6% considering themselves financially sustainable.
Would this solve fake news? No. It checks whether the text matches the content; it does not establish ground truth. But a real clip wearing a false caption is one of the most shared shapes of misinformation, and flagging it before the share button would limit it. Limited is not solved. Limited is still fewer people misled.