{
  "id": 3063656,
  "title": "Why scrolling screenshots fall apart on video feeds",
  "url": "https://urgent.news/2026/08/24/why-scrolling-screenshots-fall-apart-on-video-feeds",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-08-24T16:20:57.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/codnet_softwarecompany_7/why-scrolling-screenshots-fall-apart-on-video-feeds-ad0"
  },
  "original_language": "en",
  "account": "In attempting to fix an Android app that stitches scrolling screens into one long image, the reporter discovered two underlying defects that were causing issues when the app was pointed at social feeds. Initially, the app worked well but began capturing overlaps around video posts, leading to confusion over the reported defects. The reporter found that the measurement process could only detect one of the two defects at a time.\n\nTo address the issue, the reporter first ensured that the failure could be reproduced offline. They modified the app to write every delivered frame's grayscale data to a journal file, allowing for replay against candidate algorithms and providing a reliable way to compare different engines.\n\nBy analyzing consecutive journal frames without chaining or accumulated estimates, the reporter gained ground truth that was independent of the app's measurement. This allowed them to determine the real scroll distance between frames and the true page position of each frame. They discovered that the measurement was flawed, as it failed to account for video content and its coherent motion, which created an alternate minimum in the offset landscape.\n\nThe reporter identified the first defect: video captures skew the measurement. They built a noise fixture to isolate the issue and found that real video is structured and moves coherently, causing a second minimum in the offset landscape at the video's own motion. When video coverage exceeded the page content, the majority-of-pixels metric preferred the video's minimum, causing rows to land at the wrong page address. To fix this, the reporter suggested excluding moving rows from the measurement, but warned of potential traps in this approach.\n\nThe second defect arose from the inability of the app to detect overlapping areas. Even though the audit reported zero misplaced rows, the reporter understood that the video area's rows were assembled from different instants, causing a visible cut in the scene. To address this, the reporter proposed adding a measure to count the number of distinct source frames contributing to each animated page range. If an area is coherent, the whole range should be written from that single frame, with static rows riding along the sides to ensure proper placement.\n\nThe reporter emphasized the importance of building ground truth that does not run through the thing being tested, as shared symptoms often mask the actual issue. They recommended using fixtures that are conservative in their judgments and avoiding feedback loops that could reinforce incorrect measurements. By applying these changes, the reporter believed the app would function correctly when dealing with video content overlaying social feeds.",
  "summary": "I maintain a small Android app that stitches a scrolling screen into one long image. It worked well until people started pointing it at social feeds, and then users began reporting that captures \"overlapped\" around video posts. Chasing that report taught me it was two separate defects wearing one description, and that my measurement could only see one of them. The measurement comes first Before…",
  "key_points": [
    "App initially worked well but captured overlaps around video posts",
    "Measurement process could only detect one of two defects at a time",
    "Video captures skew measurement by creating alternate minimum in offset landscape"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}