We present Brownian motion-based sizing of individual submicron and nanopartides in liquid samples. The advantage of our approach is that particles can freely diffuse in a 10 pm thin liquid film and are therefore always within the focal depth of a low numerical aperture objective. Particles are visualized with dark-field microscopy, and the resulting diffraction-limited spots are tracked over a wide field of view of several hundred micrometers. Consequently, it is ascertained that long 2D trajectories are acquired, which leads to significantly increased partide sizing precision. The hydrodynamic diameters of metal particles with nominal sizes ranging from 70 to 200 rim in aqueous solution were determined by tracking for up to 2 min, and it was investigated if the diffusion characteristics were influenced by the proximity of substrates. This was not the case, and the estimated diameters were in good agreement with the values obtained by electron microscopy, thus validating the particle sizing principle. Furthermore, we measured a sample mixture to demonstrate the distinction of close particle sizes and performed the conjugation of a model protein (BSA) on the nanoparticle surface. An average increase in the radius of 9 nm was determined, which corresponds to the size of the BSA protein.