We perform simulations of the capabilities of the next generation Very Large Array to image stellar radio photospheres. For very large (in angle) stars, such as red supergiants within a few hundred parsecs, good imaging fidelity results can be obtained on radio photospheric structures at 38 GHz employing standard techniques, such as disk model fitting and subtraction, with hundreds of resolution elements over the star, even with just the ngVLA-classic baselines to 1000 km. Using the ngVLA Rev B plus long baseline configuration (with baselines out to 9000 km, August 2018), we find for main sequence stars within $sim$ 10 pc, the photospheres can be easily resolved at 85 GHz, with accurate measures of the mean brightness and size, and possibly imaging large surface structures, as might occur on e.g., active M dwarf stars. For more distant main sequence stars, we find that measurements of sizes and brightnesses can be made using disk model fitting to the u,v-data down to stellar diameters $sim$ 0.4 mas in a few hours. This size would include M0 V stars to a distance of 15 pc, A0 V stars to 60 pc, and Red Giants to 2.4 kpc. Based on the Hipparcos catalog, we estimate that there are at least 10,000 stars that will be resolved by the ngVLA. While the vast majority of these (95%) are giants or supergiants, there are still over 500 main sequence stars that can be resolved, with $sim$ 50 to 150 in each spectral type (besides O stars). Note that these are lower limits, since radio photospheres can be larger than optical, and the Hipparcos catalog might not be complete. Our initial look into the Gaia catalog suggests these numbers might be pessimistic by a factor few.