Objective: Electrophysiological effects underlying ECG changes during heart failure (HF) are unclear. We study such effects using an electro-physiologically detailed computer model of the action potential (AP) conduction in the heterogeneous transmural ventricular wall. Method: Single cell AP models for Purkinje fibre (PF), endocardial, midmyocardial (M) and epicardial cells were developed based on extant voltage-clamp datasets from rabbit. These models take into account experimentally observed HF-induced remodelling in the current densities of several ionic channel currents (I_Ca,L, I_to, I_Kr, I_Ks and I_K1). The single cell models were incorporated into a one-dimensional (1D) tissue strand, with inter-cellular coupling in the M cell region reduced during HF. Results: All single cell models produced APs with characteristics matching to experimental data under both control and HF conditions. HF resulted in increased transmural heterogeneity. It also reduced the amplitude of the intracellular Ca²⁺ transient by ~40% in agreement with experimental data. In the 1D tissue model, the AP conduction from the PF to the ventricular strand produced feasible pseudo-ECG waveforms. Compared to the control condition, HF increased the QT interval by 23%, and elevated the T-wave. However, its effect on QRS duration increase was negligible (< 10%). This was attributable to the 'compensatory' effect of increased upstroke velocity (~20%) of M cells due to HF that counterbalanced the reduced intercellular electrical coupling within the M-cell region. Conclusions: The model predicts only a small decrease in the AP conduction velocity – and hence, a small increase in QRS duration during HF. This suggests that experimentally observed QRS increase in HF may result from non-electrophysiological factors, such as increased thickness of the ventricular wall. However, the elevated T-wave in HF is a combined effect of greater cellular electrical heterogeneity and lower intercellular coupling.