Train chess by voice: drills you answer out loud

Voice chess apps let you play a game without touching the board: you say "knight f3" and the piece moves. Training by voice is a different thing. You answer a drill out loud, a square, a colour or a move, and the app checks you. Your hands stay free, and you cannot hover the cursor over a square while you think. On ChessTrainer.ai every drill takes voice in Chrome, Edge or Safari, and every drill also takes a tap, so a browser without a microphone loses nothing.

Playing by voice and training by voice

If you want to play games by voice, three products already do it. Verbal Chess plays on Lichess with voice alone, announces your opponent's moves and can be used without looking at the screen. VoiceChess takes moves in English, Spanish or Russian on its own site, on Lichess and, through a Chrome extension, on chess.com, and it has a set of blindfold modes. chess.com's own voice input works on the web while you hold the space bar and covers moves, draw offers, resigning and rematches.

Those are for games. Training by voice is for the part before games. A drill shows you a square and you say its name. A drill names a square and you say whether it is light or dark. An opening trainer waits for the next move of the line, and you say it. The board never needs your hands. With a mouse, most people find e4 by moving the cursor and watching where it lands. With a voice, the answer has to be in your head before you open your mouth.

How it works in the browser

The drills use the Web Speech API, which Chrome, Edge and Safari ship on desktop and on phones. You allow the microphone once. What you say is turned into text, such as "e four", and the drill checks it against the answer. Where the browser supports it, recognition runs on your device; otherwise the browser's built-in service does it, which in Chrome and Edge means the audio goes to Google's speech service under Google's privacy policy, and in Safari to Apple's. To improve recognition the site keeps about two seconds of audio for answers the recogniser got wrong, plus a small sample of correct ones, stored without your name or account and deleted after 90 days. Pro users can turn that off in Profile. In a browser without the API the drills work by tapping and the microphone button does not appear.

Chrome stops listening after a stretch of silence or about a minute of continuous speech. The drill restarts the microphone on its own, which you may notice as a short flicker of the mic indicator. If an answer fell into that gap, say it again.

What you can say

To answer	Say
A square	"e four" or "e4". The last square you say wins, so you can correct yourself.
A colour	"light" or "dark". "White" and "black" work too.
A move	"knight f3", "bishop takes e5", "castle short", "castle long", "e8 equals queen".
Hear the prompt again	"repeat" or "again".
Skip	"next" or "skip".
A hint	"hint" or "help".
Take a move back	"undo" or "take back".
End the session	"stop" or "quit".

The parser expects speech recognition to mishear chess words, so "night", "horse" and "rook" all reach the right piece, "takes" and "captures" both mean a capture, and filler words such as "please" or "move" are ignored. Say the move the way you would say it to a person across the board.

What voice is good for

Coordinates. Name the Square lights a square and waits for you to say its name; this is the drill to start with, and it is free at every level. Square Colors asks for light or dark. Both are faster by voice than by tapping, because the answer is one word and your eyes never leave the board.

Openings. Say each move of the line and the trainer plays the reply. Saying "bishop c4" rather than dragging a bishop makes you name the square, which is the part that sticks. See the openings guide for how the review dates work.

Blindfold and visualization. When the pieces are hidden, voice is the only honest input, since tapping a hidden square lets you count files with a finger. Blindfold chess benefits covers what the training is for.

Anywhere your hands are busy. Cooking, walking, a treadmill, a child asleep on your arm. A coordinates level takes under a minute and needs one word per square.

Where it fails

• Noisy rooms. Recognition drops fast with a television or a second conversation in range. Earbuds with a microphone help more than speaking louder.
• Letters that sound alike. B, D, E and G are one consonant apart, and some accents push them together. If a letter keeps coming back wrong, say the square a second time more slowly, or tap it and move on.
• A browser without the API. Some browsers, including the in-app browsers inside social apps, have no speech recognition. Open the page in Chrome, Edge or Safari, or tap.
• A denied microphone. If you dismissed the permission prompt, the mic button shows an error and how to allow it again in your browser.

Accessibility

If you cannot use a mouse or a touchscreen, every drill on the site can be run by voice from start to finish: the answers, "next", "repeat", "hint" and "stop". The drills speak their prompts when sound is on, so with voice answers a coordinates or colour session works without looking at the screen. If a drill cannot be completed by voice on your setup, say so through the feedback form in the app and it will be treated as a bug.

To try it, open Name the Square, press the microphone button, allow the microphone, and say the name of the lit square.