Considering the relatively high precision that will be reached by future observatories, it has recently become clear that one dimensional (1D) atmospheric models, in which the atmospheric temperature and composition of a planet are considered to vary only in the vertical, will be unable to represent exoplanetary transmission spectra with a sufficient accuracy. This is particularly true for warm to (ultra-) hot exoplanets because the atmosphere is unable to redistribute all the energy deposited on the dayside, creating a strong thermal and often compositional dichotomy on the planet. This situation is exacerbated by transmission spectroscopy, which probes the terminator region. This is the most heterogeneous region of the atmosphere. However, if being able to compute transmission spectra from 3D atmospheric structures (from a global climate model, e.g.) is necessary to predict realistic observables, it is too computationally expensive to be used in a data inversion framework. For this reason, there is a need for a medium-complexity 2D approach that captures the most salient features of the 3D model in a sufficiently fast implementation. With this in mind, we present a new open-source documented version of Pytmosph3R that handles the computation of transmission spectra for atmospheres with up to three spatial dimensions and can account for time variability. Taking the example of an ultrahot Jupiter, we illustrate how the changing orientation of the planet during the transit can allow us to probe the horizontal variations in the atmosphere. We further implement our algorithm in TauREx to allow the community to perform 2D retrievals. We describe our extensive cross-validation benchmarks and discuss the accuracy and numerical performance of each model.< Réduire