We model the large-scale linear galaxy bias b(g)(x, z) as a function of redshift z and observed absolute magnitude threshold x for broad-band continuum emission from the far-infrared to ultraviolet, as well as for prominent emission lines, such as the H alpha, H beta, Ly a, and [O II] lines. The modelling relies on the semi-analytic galaxy formation model GALFORM, run on the state-of-the-art N-body simulation SURFS with the Planck 2015 cosmology. We find that both the differential bias at observed absolute magnitude x and the cumulative bias for magnitudes brighter than x can be fitted with a five-parameter model: b(g)(x, z) = a + b(1 + z)(e)(1 + exp [(x - c)d]). We also find that the bias for the continuum bands follows a very similar form regardless of wavelength due to the mixing of star-forming and quiescent galaxies in a magnitude-limited survey. Differences in bias only become apparent when an additional colour separation is included, which suggest extensions to this work could look at different colours at fixed magnitude limits. We test our fitting formula against observations, finding reasonable agreement with some measurements within 1 sigma statistical uncertainties, and highlighting areas of improvement. We provide the fitting parameters for various continuum bands, emission lines, and intrinsic galaxy properties, enabling a quick estimation of the linear bias in any typical survey of large-scale structure.