Chess notation in subtitles
Chess commentary has three natural written forms, and a caption file should pick one and stay with it. Chesscribe renders all three from the same transcript.
- Spoken keeps the commentator’s words and only fixes the chess: “knight to e4”. Best for verbatim captions and accessibility.
- Clean drops filler such as “to” and normalises captures and checks: “knight e4”, “bishop takes c6”. Best for most YouTube lessons.
- Notation writes standard algebraic notation: “Ne4”, “Bxc6”, “O-O-O”. Best for study content and on-screen graphics, and the source for PGN export.
| Whisper hears | Spoken | Clean | Notation |
|---|---|---|---|
| knight to e four | knight to e4 | knight e4 | Ne4 |
| bishop takes on see six | bishop takes on c6 | bishop takes c6 | Bxc6 |
| queen h five check mate | queen h5 checkmate | queen h5 checkmate | Qh5# |
| castles queen side | castles queenside | castles queenside | O-O-O |
| e takes d five | e takes d5 | e takes d5 | exd5 |
| e eight promotes to a queen | e8 promotes to a queen | e8 =Q | e8=Q |
| night b d two | knight b d2 | knight b d2 | Nbd2 |
House rules the grammar follows
- Squares are always lower-case: e4, not E4.
- Piece names are lower-case mid-sentence; speech models capitalise them at random and that noise is removed.
- “Check mate” becomes one word; “king side” and “queen side” likewise.
- Ambiguous words such as “for”, “to” and “see” only turn into ranks and files when a piece or a capture sits next to them, so ordinary sentences are left alone.