your real board, live
point your phone at the board and the position is read as it is on the table, in real time.
beta · waitlist open
point your phone at a real board.
vision reads the pieces on a real, three-dimensional chessboard through your camera and converts the position to FEN in real time, so you can import it into zochess analysis and analyse it.
vision is in beta. leave your email and we will let you know when it opens.
$ vision --why█
point your phone at the board and the position is read as it is on the table, in real time.
the FEN you get imports into zochess analysis, so a position from your table can be analysed like any online game.
we are also working on converting over-the-board video into PGN, so a whole game can be captured and not just a position.
$ vision --preview█
hover a box to see its piece, confidence and square
r2k3r/2qn1pQp/p2pp3/3b1N2/1p3P2/3B4/PPP3PP/2KRR3 w - - 0 1█
example scan: real output of the zochess vision detector on a synthetic test image. every piece is found, and the position matches the ground truth exactly.
$ vision --when it helps█
a position on a real board is easy to lose and slow to copy. digitizing it turns it into something you can analyse, save and share.
point your phone at the board and it is saved as FEN, with no retyping.
scan the board and check it with an engine before the pieces are put away.
scan mid-game and you always have a copy, even if the table gets bumped.
set up a position on a real board and everyone gets it on their phone.
share the exact position with someone online, without describing it.
$ vision --features█
reads your board through the camera as you look at it.
recognises the pieces on a real, three-dimensional chessboard.
converts the position to FEN, the standard notation that chess tools understand.
planned: turn a video of a whole over-the-board game into a PGN.
$ vision --how it works█
place your phone so the board is in view.
vision recognises the pieces and converts the position to FEN in real time.
import the FEN into zochess analysis to see the evaluation and the best moves.
$ vision --faq█
an app that reads a real chessboard through your phone's camera and converts the position to FEN in real time.
placing 20 or more pieces on an analysis board takes minutes, and it is easy to put one on the wrong square. a scan takes a single look.
a short line of text that describes a chess position. lichess, chess.com and every chess engine understand it.
a photo shows where the pieces are, not whose move it is or who can still castle. vision assumes white to move and no castling, and you can change that when you import the position.
tap a square to correct it before you save. if the board is turned the wrong way, one tap rotates it.
it works on most classic boards today. we are still improving it for other styles, and more real photos are what make it better.
on our synthetic test images the board was found every time, 75 of 99 positions were exactly right and 96.8% of squares were correct. we are still measuring accuracy on real photos.
yes. scans are uploaded to your account by default, so your history is kept wherever you sign in. they are private, and we do not use them to train models unless you choose to (that option is not available yet). you can delete any scan, or your account, at any time.
scan real positions and correct any mistakes. every scan and correction makes the model better, especially on boards it has not seen much.
we are working on converting over-the-board video into PGN. it is planned and not available yet.
vision is in beta and has not been released yet. join the waitlist to hear when it opens.
in beta now. join the waitlist for early access and help us improve the detector.