YouTube quality — when a video Gist feels off

Gist works from the words, not the pictures. A clean transcript gives the best Gist; with no captions it falls back to the title and description, so quality varies.

Gist reads the words, not the pictures — the transcript when there is one, the title and description when there is not.

When a transcript exists, the quality of the Gist tracks the quality of that transcript:

  • Clean uploader-provided captions — best results.
  • Auto-generated captions in clear English — works.
  • Auto-generated captions on a heavily-accented or low-audio video — mixed.
  • No captions — best-effort only. Gist falls back to the title and description, so a short, well-described video may still produce a usable Gist, but expect lower quality. A long video with no captions and a thin description will not.

Videos that work

  • Talks and conference videos. Clear speech, structured argument, captioned.
  • Lectures and courses. Same.
  • Interviews and podcasts uploaded to YouTube. Transcript-based, captioned, layers map cleanly.
  • Explanation videos. The layered read matches the video's own structure.

Videos that don't

  • Music videos. No argument to layer.
  • Short-form vlogs. Too little content to layer.
  • Visual-heavy content — cooking shows, sports highlights. The transcript misses most of what is happening.

What to try

  • Check captions. On YouTube, hit CC. Can you read them? If yes, Gist should too.
  • Use the full video URL, not a timestamped one. Gist reads the whole transcript; deep-linking does not change what it reads.
  • For live streams, wait until the stream is over and uploaded with a final transcript.

When to contact support

A video with good captions and clear content that consistently fails or produces a wrong Gist is a bug. Send it via contact support:

  • The YouTube URL.
  • A short description of what is off — missing context, wrong thesis, layers swapped.
  • Whether captions are on.

What is next