Video knowledge

A transcript is not the same asset as the video it came from, and treating it as one is where most video pipelines lose their value. Timestamps are what make a claim checkable later; strip them to save space and you have produced text that nobody can verify against the source. Language detection matters for the same reason, since an auto-translated line carries a confidence the original did not. The guides here cover the full path from a public URL to searchable, attributable text: choosing between captions and generated transcripts, handling multi-speaker audio, keeping metadata attached through export, and deciding when translation helps versus when it quietly introduces an interpretation you did not intend to publish.