What it cleans up
A copied YouTube transcript carries two kinds of noise around the words. The converter removes both:
- Timestamps like
0:08are dropped (or kept as a compact inline prefix, your choice). - Accessibility labels get glued to the text with no space, so
0:08plus its "8 seconds" label plus the words arrives as0:088 secondscouple ideas. The converter separates and removes the label, leaving justcouple ideas. - Fragmented cues are reflowed into paragraphs, so the result reads like prose instead of two-word caption snippets.
The words themselves are left untouched, including filler like "um" and repeated words, because removing them would change what was said.