Vessel-selective dynamic angiograms provide a wealth of useful information about the anatomical and functional status of arteries, including information about collateral flow and blood supply to lesions. Conventional x-ray techniques are invasive and carry some risks to the patient, so non-invasive alternatives are desirable. Previously, non-contrast dynamic MRI angiograms based on arterial spin labeling (ASL) have been demonstrated using both spoiled gradient echo (SPGR) and balanced steady-state free precession (bSSFP) readout modules, but no direct comparison has been made, and bSSFP optimization over a long readout period has not been fully explored. In this study bSSFP and SPGR are theoretically and experimentally compared for dynamic ASL angiography. Unlike SPGR, bSSFP was found to have a very low ASL signal attenuation rate, even when a relatively large flip angle and short repetition time were used, leading to a threefold improvement in the measured signal-to-noise ratio (SNR) efficiency compared with SPGR. For vessel-selective applications, SNR efficiency can be further improved over single-artery labeling methods by using a vessel-encoded pseudo-continuous ASL (VEPCASL) approach. The combination of a VEPCASL preparation with a time-resolved bSSFP readout allowed the generation of four-dimensional (4D; time-resolved three-dimensional, 3D) vessel-selective cerebral angiograms in healthy volunteers with 59 ms temporal resolution. Good quality 4D angiograms were obtained in all subjects, providing comparable structural information to 3D time-of-flight images, as well as dynamic information and vessel selectivity, which was shown to be high. A rapid 1.5 min dynamic two-dimensional version of the sequence yielded similar image features and would be suitable for a busy clinical protocol. Preliminary experiments with bSSFP that included the extracranial vessels showed signal loss in regions of poor magnetic field homogeneity. However, for intracranial vessel-selective angiography, the proposed bSSFP VEPCASL sequence is highly SNR efficient and could provide useful information in a range of cerebrovascular diseases. © 2016 The Authors. NMR in Biomedicine published by John Wiley & Sons Ltd.