[en] Context: Detection of solar gravity modes remains a major challenge to our understanding of the inner parts of the Sun. Their frequencies would enable the derivation of constraints on the core physical properties, while their amplitudes can put severe constraints on the properties of the inner convective region. Aims: Our purpose is to determine accurate theoretical amplitudes of solar g modes and estimate the SOHO observation duration for an unambiguous detection of individual modes. We also explain differences in theoretical amplitudes derived from previous works. Methods: We investigate the stochastic excitation of modes by turbulent convection, as well as their damping. Input from a 3D global simulation of the solar convective zone is used for the kinetic turbulent energy spectrum. Damping is computed using a parametric description of the nonlocal, time-dependent, convection-pulsation interaction. We then provide a theoretical estimation of the intrinsic, as well as apparent, surface velocity. Results: Asymptotic g-mode velocity amplitudes are found to be orders of magnitude higher than previous works. Using a 3D numerical simulation from the ASH code, we attribute this to the temporal-correlation between the modes and the turbulent eddies, which is found to follow a Lorentzian law rather than a Gaussian one, as previously used. We also find that damping rates of asymptotic gravity modes are dominated by radiative losses, with a typical life time of 3 × 10[SUP]5[/SUP] years for the ell=1 mode at nu=60 muHz. The maximum velocity in the considered frequency range (10-100 muHz) is obtained for the ell=1 mode at nu=60 muHz and for the ell=2 at nu=100 muHz. Due to uncertainties in the modeling, amplitudes at maximum i.e. for ell=1 at 60 muHz can range from 3 to 6 mm s[SUP]-1[/SUP]. The upper limit is too high, as g modes would have been easily detected with SOHO, the GOLF instrument, and this sets an upper constraint mainly on the convective velocity in the Sun.