Working on reconstructing a virus with a box size of 1000. Ab-initio (no symmetry, single classs) generates a great initial model, but it takes a considerable amount of time to finish (~8 hrs - too long for cryoSPARC!). It would be great if one could bin their data by 2 or 4 for from inside the cryoSPARC to generate an initial model and then use unbinned for refinement.
Your suggestion worked - it shortened the time it takes to run significantly. After carefully looking I also found the parameter “Refinement box size (Voxels)” which should take care of binning for refinement.