<p>Earthquake-resistant design has conventionally relied on force-based methods in design codes, while displacement-based approaches remain under development. Both approaches use peak response measures such as acceleration or displacement but do not capture the cumulative effects of seismic shaking. These effects can be represented through the total input energy, which provides a more comprehensive measure of seismic demand and serves as primary step towards energy-based design. While energy-based ground motion models (eGMM) for Peninsular India and Himalaya are developed, no such model existed for ground motions in Indo Burma region (IBR) despite its significant seismic hazard. This study develops an artificial neural network (ANN)-based ground-motion model for the IBR events. The model predicts the input energy equivalent velocity (V_ea) using the NGA subduction database and regionalises for IBR events. Eight input variables are used, including magnitude, distance, site condition, depth, and fault mechanism, with outputs defined as V_ea at 23 natural periods from 0.01 to 5s. Model performance is evaluated through residual analysis, parametric studies to confirm physical trends, and Shapley Additive Explanations (SHAP) to assess variable importance. The proposed model is valid for moment magnitudes between 5.0 and 8.5 for IBR events (4.0–9.5 in NGA subduction), Joyner–Boore distances up to 500 km, and site conditions with V_s30 values up to 2000 m/s, covering the full IBR site classification range. The results demonstrate that the ANN-based eGMM provides reliable prediction of the energy spectrum and offers a practical tool for advancing energy-based earthquake-resistant design in the region.</p>