danielmiessler/Fabric v1.4.520

on GitHub

Changes

PR #2101 by shaun0927 and ksylvan: fix(youtube): preserve ffmpeg frame timing in visual OCR cues

• Fixes YouTube visual OCR cues so timestamps come from real ffmpeg frame timing instead of the OCR result index.
• Parses pts_time from ffmpeg showinfo output and formats cue ranges from actual frame times, with a safe fallback.
• Adds regression coverage for the timestamp parser, cue-range helper, and the end-to-end visual extraction path.