The 'Bloated Stars Scenario' proposes that AGN broad-line emission originates in the winds or envelopes of bloated stars (BSs). Its main advantage over broad-line region (BLR) cloud models is the gravitational confinement of the gas, and its major difficulty the large estimated number of BSs and the resulting high collisional and evolutionary mass-loss rates. Previous work on this model did not include full calculations of the spectrum and used very simplified stellar distribution functions. Here we calculate the emission-line spectrum by applying a detailed numerical photoionization code to the wind and by assuming a detailed nucleus model (Murphy, Cohn & Durisen). We study a wide range of wind structures for a QSO model, with L(ion) = 7x10(45) erg s(-1), M(bh) = 8x10(7) M. and a stellar core density of 7x10(7) M, pc(-3). The BSs emit a spherically symmetric wind, whose size and boundary density are determined by Various processes: Comptonization by the central continuum source, calculated here self-consistently; tidal disruption by the black hole; and the limit set by the wind's finite mass. We find that the emission spectrum is mainly determined by the conditions at the boundary of the line-emitting fraction of the wind, rather than by its internal structure. Comptonization results in a very high ionization parameter at the boundary, which produces an excess of unobserved broad high-excitation forbidden lines. The finite mass constraint limits the wind size, increases the boundary density and thus improves the results. Slow, decelerating, mass-constrained flows with high gas densities (10(8) to 10(12) cm(-3)) are as successful as cloud models in reproducing the overall observed line spectrum. The MgrII lambda 2798 and Nv lambda 1240 lines are, however, underproduced in our models. We adjust the number of BSs so as to obtain the observed EW(Ly alpha) and find that only similar to 5 x 10(4) BSs with dense winds are required in the inner 1/3 pc. This small fraction approaches that of supergiants (SGs) in the solar neighbourhood and the calculated mass loss is consistent with the observational constraints. The required number of BSs and their mass-loss rates are very sensitive functions of the wind density structure. SG-like BSs are ruled out by their high mass loss. BSs with dense winds can reproduce the BLR line spectrum and be supported by the stellar population without excessive mass-loss and collisional destruction rates. The question whether such hitherto unobserved stars actually exist in the BLR remains open.
