Résumé : Light (anti-) nuclei are a powerful tool both in collider physics and astrophysics. In searches for new and exotic physics, the expected small astrophysical backgrounds at low energies make these antinuclei ideal probes for, e.g., dark matter. At the same time, their composite structure and small binding energies imply that they can be used in collider experiments to probe the hadronisation process and two-particle correlations. For the proper interpretation of such experimental studies, an improved theoretical understanding of (anti-) nuclei production in specific kinematic regions and detector setups is needed. In this work, we develop a coalescence framework for (anti-) deuteron production which accounts for both the emission volume and momentum correlations on an event-by-event basis. This framework goes beyond the equal-time approximation, which has been commonly assumed in femtoscopy experiments and (anti-) nucleus production models until now. Using PYTHIA~8 as an event generator, we find that the equal-time approximation leads to an error of O(10%) in low-energy processes like $\Upsilon$ decays, while the errors are negligible at LHC energies. The framework introduced in this work paves the way for tuning event generators to (anti-) nuclei measurements.