This study aims to investigate ambient vibration based modal analysis and cable tension estimation with Bayesian approaches, associating with an ambient vibration testing in a cable-stayed bridge. Firstly, a Bayesian fast Fourier transform (FFT) is introduced to the target bridge for operational modal analysis using ambient vibration data. Considering a mixing effect of local modes and global modes in the vibration measurements of cables, the identification is implemented in three scenarios: global modal analysis without data of cables, global modal analysis with data of cables, local modal analysis of each cable. The effects of involving measurements of cables into the global modal analysis of the whole bridge are discussed, and the identification uncertainty of modal frequencies in each scenario is investigated. Then, considering the identified modal frequencies of multiple modes in each cable, a Bayesian cable tension estimation framework including a Bayesian linear regression (BLR) is proposed, with the ability of simultaneously estimating the cable tension and flexural rigidity in a probabilistic way. The estimation is conducted with the identification results by both ambient vibration and hammer test. The results showed different accuracy and uncertainty in the estimation of cable tension and flexural rigidity, as well as among different cable conditions.