Disordered fibre networks are ubiquitous in nature and have a wide range of industrial applications as novel biomaterials. Predicting their viscoelastic response is straightforward for affine deformations that are uniform over all length scales, but when affinity fails, as has been observed experimentally, modelling becomes challenging. Here we introduce a numerical methodology to predict the steadystate viscoelastic spectra and degree of affinity for disordered fibre networks driven at arbitrary frequencies. Applying this method to a peptide gel model reveals a monotonic increase of the shear modulus as the soft, non-affine normal modes are successively suppressed as the driving frequency increases. In addition to being dominated by fibril bending, these low frequency network modes are also shown to be delocalised. The presented methodology provides insights into the importance of non-affinity in the viscoelastic response of peptide gels, and is easily extendible to all types of fibre networks. [5]. Nature employs protein fibre networks in the multi-functional cellular cytoskeleton [6,7]. The mechanical stiffness of fibre networks is often central to their function, and although static properties come under most scrutiny, they often exist in dynamic environments subject to temporally-varying mechanical loads, including the cytoskeleton of motile cells [7], and scaffolds for tendon and ligament regeneration, where habitual loading propagating through the network influences the viability of embedded stem cells [8][9][10]. Understanding the dynamical network response is essential to design novel materials with properties suited for such situations.A key modelling challenge is to determine the degree to which the deformation is affine [11], i.e. uniform over all relevant length scales; see Fig. 1. If affinity holds, extrapolating the macroscopic response from a putative microstructure is straightforward, and a range of thermal and athermal affine models for fibre networks have been developed [12,13]. When affinity fails, however, as experimentally observed over broad parameter ranges [14][15][16][17], it is necessary to determine the microscopic deformation field, which typically requires numerical solution for explicit network realisations. This has thus far been limited to the elastic plateau amenable to energy minimization algorithms [18][19][20][21], or computationally-intensive particle methods that only access short times [22,23]. Without a more general understanding of fibre networks dynamics, we lack the capability to predict potentially large changes in viscoelastic properties over experimentally relevant time scales.Here we present a methodology which allows the numerical calculation of the viscoelastic spectra for any type of disordered fibre network driven at arbitrary oscillation frequencies. The method is based on normal modes which ensures linear response, and since no thermal effects or crosslink dynamics are included by construction, all measured variation in affinity and viscoelasticity can be as...