In mammals, transcription was observed to occur predominantly in bursts, resulting from short and intense periods of gene transcription interspersed by longer silent periods. In order to elucidate the causes and the consequences of discontinuous transcription in mammals, we further extended a probabilistic and computational framework to model single-cell time-lapse recording of luminescent reporter based on stochastic gene expression models that describe the basic processes of gene activation, transcription, translation, and degradation of mRNA and proteins. Motivated by the recent finding of a refractory period in gene reactivation, we performed inference in promoter cycle models, where the gene activity is modeled by one active state and multiple sequential inactive states describing the promoter progression toward activation, and we characterized the rate limiting processes causing the refractory period. Our analysis demonstrated that the kinetic structure of the cycle is gene-specific. Several rate limiting steps (5 on average) are required to model the silent period of endogenous promoters while minimal synthetic promoters overall exhibit a simplified promoter cycle structure characterized by a reduced number of steps (3 on average) and longer silent period. Nevertheless, the promoter architecture might not solely be responsible for such an observed difference, as the synthetic promoters were all sharing similar chromatin context (same insertion locus). Based on the inferred kinetics, we investigate the implications on the transcriptional noise. It followed from the inferred bursting kinetics, that intrinsic noise is dominated by the fluctuations in the cycle, and globally maintained over the wide range of expression covered by the clones (100-fold). With minor modifications of the developed probabilistic framework, we investigated how the bursting kinetics of a mammalian gene (Ctgf) are modulated in response to two different induction pathways. Both stimuli resulted in a fast and acute increase in burst sizes. Whereas TGF-b1 induced a sustained increase of transcription rate and prolonged transcriptional activation, serum stimulation led to a large and temporally tight first transcriptional burst, followed by a refractory period in the range of hours. Those observations illustrated how different stimuli can result in kinetically distinct transcriptional responses of the same gene.