The application of genetically encoded fluorophores for microscopy has afforded one of the biggest revolutions in the biosciences. Bioluminescence microscopy is an appealing alternative to fluorescence microscopy, because it does not depend on external illumination, and consequently does neither produce spurious background autofluorescence, nor perturb intrinsically photosensitive processes in living cells and animals. The low quantum yield of known luciferases, however, limit the acquisition of high signal-noise images of fast biological dynamics. To increase the versatility of bioluminescence microscopy, we present an improved low-light microscope in combination with deep learning methods to increase the signal to noise ratio in extremely photon-starved samples at millisecond exposures for timelapse and volumetric imaging. We apply our method to image subcellular dynamics in mouse embryonic stem cells, the epithelial morphology during zebrafish development, and DAF-16 FoxO transcription factor shuttling from the cytoplasm to the nucleus under external stress. Finally, we concatenate neural networks for denoising and light-field deconvolution to resolve intracellular calcium dynamics in three dimensions of freely moving Caenorhabditis elegans with millisecond exposure times. This technology is cost-effective and has the potential to replace standard optical microscopy where external illumination is prohibitive.