> I am also interested in main effect of conditions and in comparing condition 1 among the subjects in group 1 and condition 3 among subjects in group 2.
You could operationally compare the two conditions with a two-sample t-test (using 3dttest++), but from the modeling perspective it's really a bad model and a poor experimental design. Ideally you want to have both conditions available for the two groups.
> I would like to know where I could find voxel sizes, then I can use them to set the Clustersize in the AFNI GUI (multiple comparisons correction) to see significant clusters.
You can try 3dttest++ -Clusterize or -ETAC. Read the help documentation for details.
Gang