Course audio should have transcripts

[jiangxue-111]

Audio is useful, but transcripts are necessary for searching, silent review, accessibility, and checking technical terms. Even if word-level sync is hard, paragraph-level transcript would already help.

[Gelisi]

Thanks, this makes sense. A practical first step could be to store the generated narration text as a transcript for each course section, even before doing word-level synchronization. Later we can consider paragraph-level highlighting during playback, which would also help with silent review or copying key explanations into notes.