Skip to main content
Video understanding helps with short clips, course recordings, product demos, surveillance summaries, and media review. Depending on the model, you can send video directly or process frames and captions separately.

​
Common tasks

  • Clip summary: describe main content, key moments, and conclusions.
  • Visual QA: ask questions about a frame or time range.
  • Action recognition: analyze actions, process steps, or event order.
  • Subtitle cleanup: combine visual context with existing subtitles to fix wording, timing, and continuity.

​
Integration note

If a model does not support direct video input, extract key frames for image understanding, then summarize frame notes, captions, and timestamps with a text model.

​
Next steps