Turn a Lichess game into a narrated review video

A developer used Claude Code to blend engine analysis with spoken notes and make a commented video of a chess game.

Done with Claude Code· Advanced

What they did
They recorded themselves talking through a 15+10 rapid game, then shared the Lichess game link and an audio file with Claude Code. The agent transcribed the recording locally with whisper.cpp, lined up the spoken remarks with moves using PGN clock times, and ran Stockfish over the whole game. When it found errors, subagents asked the engine about each one, then a verification pass checked the output before the agent built a storyboard and narrated video.
How it went
It created an annotated PGN, an HTML board view, and a 6:49 video with subtitles based on the author’s actual game. The author said mistakes are uncommon, but the system can still make them.
Tips
This workflow takes about an hour, so it fits slower games where there was time to think. It is less useful for blitz, where there may not be enough spoken reflection to match cleanly to the moves.

Try it yourself with Claude Code

Review my chess game from [Lichess game URL] using Stockfish and my own commentary in [audio file or text file]. Transcribe the commentary, match it to the moves using the game clock times, and identify the main inaccuracies, missed tactics, and best alternatives. Then draft an annotated move list and a storyboard for a narrated video, and stop before exporting or sharing anything irreversible until I approve the final cut. After my approval, produce the finished video with subtitles and also give me a short summary of the key positions covered.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on github.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.