Changelog
Every user-visible change, dated. If a checker ever gives a wrong result, the fix and the affected period will be recorded here — see our incident playbook commitments on the About page.
21 November 2026 — the full toolkit
- Video Analyzer upgraded: true loudnorm LUFS measurement in the browser (replacing the estimate for files up to 500 MB), face presence in the opening frames, loop-point check for short videos, provenance mini-scan, segment mode for files over the 1 GB / 60-minute gate, and a stall watchdog.
- New tools: Pre-Upload Checklist (with a transparent mechanical title score), Song Identifier, Channel Health Report, Own-Channel Repetition Scan, Description Link Checker, Duplicate Video Check, Screenshot Authenticity Clues.
- EXIF & Provenance Viewer now extracts a C2PA manifest's own claims (generator, declared actions, source type, names in the certificate text) — labelled extracted, never verified.
- Song Copyright Risk wizard now branches by platform (YouTube vs Instagram/Facebook vs TikTok — the same song is treated differently on each).
- Real Quality Checker: thresholds recalibrated against real files (a native 4K master and a deliberate 1080p→4K upscale control). A spectral (FFT) signal was implemented, measured, found not to separate real files, and removed before shipping — the validated round-trip method stands alone.
- Ten creator guides published, including the EU AI Act for creators and a German AI-labelling page.
- The loudness engine (~31 MB) is now cached by a service worker after first use.
November 2026 — first build
- First 20 tools built: Video Analyzer, Region Checker, Monetization Checker, Metadata Viewer, Thumbnail Checker, LUFS Checker, Real Quality Checker, EXIF & Provenance Viewer, AI Label Checker, Song Copyright Risk, Face-Swap Check (engine pending), Originality Audit, Subtitle Checker, Privacy Scan, Sponsor-Ready Report, Claim Dispute Helper, Disclosure Checker, TikTok FYF Self-Check, Yellow-Icon Check, Appeal Checklist.
- Honesty standard set on day one: evidence + confidence wording everywhere; no shadowban verdicts, no "AI probability" scores, no guarantees — and every tool carries a "what this tool cannot tell you" box.