Upgrades to current gravitational wave detectors for the next observation run and upcoming third-generation observatories, like the Einstein telescope, are expected to have enormous improvements in detection sensitivities and compact object merger event rates. Estimation of source parameters for a wider parameter space that these detectable signals will lie in, will be a computational challenge. Thus, it is imperative to have methods to speed-up the likelihood calculations with theoretical waveform predictions, which can ultimately make the parameter estimation faster and aid in rapid multi-messenger follow-ups. In this work we study auto-encoder models for gravitational waveform generation by adopting the best-performing architecture of Liao & Lin (2021) to approximate aligned-spin SEOBNRv4 inspiral-merger-ringdown waveforms. Our parameter space consists of four parameters, [$m_1$, $m_2$, $χ_1(z)$, $χ_2(z)$]. The masses are uniformly sampled in $[5,75]\,M_{\odot}$ with a mass ratio limit at $10\,M_{\odot}$, while the spins are uniform in $[-0.99,0.99]$. Our model is able to generate $10^3$ waveforms in $\sim 0.1$ second at an average speed of about 50 microsecond per waveform on a GPU. This is about 4 orders of magnitude faster than the native SEOBNRv4 implementation, and 2--3 orders of magnitude faster than existing non-machine-learning accelerated waveform variants. The median mismatch for the generated waveforms in the test dataset is $\sim10^{-2}$, with better performance in a restricted parameter space of $χ_{\rm eff}\in[-0.80,0.80]$. The latent sampling error of our model can be quantified at a median mismatch standard deviation of $4\times10^{-3}$. Although the accuracy of our model does not enable full production-use yet, the model could be useful wherever high-volume of approximate theoretical waveforms are required, for instance, for rapid sky localization.