This paper presents and evaluates a new security primitive in the form of non-transferable ``visual secrets''. We show how they can be used in the design of voting systems. More specifically, we introduce a receipt-free low-tech visually verifiable boardroom voting system which is built for simplicity and can serve as a teaching tool to introduce people to verifiable voting. Visual secrets rely on the pre-semantic treatment of images in the human brain. After being shown an image for a limited time, users can recognise it in a larger set (79% recognition compared to an expected baseline of 5%). However, they cannot reliably communicate to someone else exactly how to do so (whether voluntarily or through coercion). Indeed, no assessor managed to accurately identify the images from the description given by the subjects. We then introduce a boardroom voting system based on this primitive. The voter receives a ballot consisting of a single picture, votes by folding it horizontally or vertically and casts it. If the voter is coerced into describing their ballot, several ballots are likely to correspond to the description. When all ballots are revealed, the voter can check with a glance that their ballot is present and folded correctly. This gives them the opportunity to detect error or fraud, although they cannot prove that fraud happened (the limited dispute resolution mechanism mainly focuses on incentives). The design makes use of textured paper to provide both accessibility for the blind and improved usability for all users.