In single-molecule biophysics and super-resolution microscopy, fluorescent probes are routinely localized with nanometer precision in images taken, e.g., with an EMCCD camera. In such images, an isolated probe images as a diffraction-limited spot of light which was formed by a finite number of photons. The probe’s coordinates are estimated from the recorded camera intensities in the spot, and the error on this estimate, the localization error, is given by a mathematical formula that depends on the number of photons in the spot. Translation of measured intensities to photon numbers requires a calibration of the camera for the specific setting with which it is used. Here we show how this can be done post festum from just a recorded image. We demonstrate this (i) theoretically, mathematically, (ii) by analyzing images recorded with an EMCCD camera, and (iii) by analyzing simulated EMCCD images for which we know the true values of parameters. In summary, our method of calibration-on-the-spot allows calibration of a camera with unknown settings from old images on file, with no other info needed. Consequently, calibration-on-the-spot also makes future camera calibrations before and after measurements unnecessary, because the calibration is encoded in recorded images during the measurement itself, and can at any later time be decoded with calibration-on-the-spot.