Since it’s K3 TIFF data, chances are it was acquired with a separate gain reference. Micrographs will probably look streaky and/or show defects which impact all stages of processing. If you can find the gain reference that would probably be quickest. However, if that is not possible there are ways around it.
How many micrographs do you have? The K3 is a great detector but the gain drift can be dramatic. If you’ve got 10,000+ micrographs a single gain reference may not be good enough for all micrographs to be correctly… er… corrected.
If you don’t have RELION installed and working already, your best bet will be to download the pre-compiled cisTEM binaries and use the sum_all_tiff_files application to generate a sum which can be used as a gain reference.
Depending on how many micrographs you have, you can either sum all of them, or sum all of them in different sets (you will need to symlink them to different directories and run sum_all_tif_files in each directory - remember to name the gain references differently!) but personally I’d usually do it in groups of 500-2000 depending on a variety of factors (total micrographs, single acquisition period, number of frames per micrograph, what a couple of test runs for gain generation looks like…) you don’t need to sum all micrographs - just enough to get a good overall sum.
If you do have RELION installed, you can import the micrographs into a .star file and then use relion_estimate_gain to generate a gain reference.