High angular resolution images obtained with a hypertelescope can strongly constrain the radiative-hydrodynamics simulation of red supergiant (RSG) stars, in terms of intensity contrast. granulation size and temporal variations of the convective motions that are visible on their surface. The characterization of the convective pattern in RSGs in crucial to solve the mass-loss mechanism which contributes heavily to the chemical enrichment of the Galaxy. We show here how the astrophysical objectives and the array configuration are highly dependent to design a hypertelescope. For a given field of view and a given resolution, there is a trade-off between the array geometry and the number of required telescopes to optimize either the (u,v) coverage (to recover the intensity distribution) or the dynamic range (to recover the intensity contrast). To obtain direct snapshot images of Betelgeuse with hypertelescope, a regular and uniform layout of telescopes is the best array configuration to recover the intensity contrast and the distribution of the large and small granulation cells, but it requires a huge number of telescopes (several hunderds of thousands). An annular configuration allows a reasonable number of telescopes (lower than one hundred) to recover the spatial structures but it provides a low-contrast image. Concerning the design of a pupil densifier to combine all the beams, the photometric fluctuations are not critical (Delta photometry <50%) contrary to the residual piston requirements (OPD < lambda/8) which requires the development of an efficient cophasing system to fully exploit the imaging capability of a hypertelecope.