Skip to main content

Overview

In AI insights, open Visual summary (Beta). Screenshot analysis now picks frames from the picture itself, not even time slices. Video understanding still reads the whole clip. The card notes that online videos can be analyzed without downloading first (local files, Bilibili, or YouTube). Xiaohongshu image notes can also be summarized in one tap and sent through the same keyframe analysis.
Video understanding mode in visual summary
Model picker for visual analysis

Pain Point

I want AI to see charts and demos in a YouTube or Bilibili video, but the web app used to download the file first. Long videos meant sitting through minutes of wait.

Key Features

  • Screenshot analysis: smart frames from the picture, not a fixed interval
  • Video understanding: watch the full clip; online links skip the download wait
  • Local files, Bilibili, YouTube; Xiaohongshu image notes can summarize and run keyframe analysis
  • Works with custom settings, regenerate, and chapters

Highlights

On the web you can paste a link and analyze the picture, not only the captions.

Use Cases

  • Tutorials where the important step is on screen
  • Talks with charts or a whiteboard
  • Long videos you do not want to download first

Tips

  • For direct online analysis, pick a Gemini (Google) model in the list; other models are listed too
  • Use a normal summary when captions are enough; turn on video understanding when the picture matters

Keyframe Screenshot Analysis

Chapter Deep Reading

Chat Agent Mode


Upgrade to Pro: more visual analysis runs → Pricing