Sequencing cell-free DNA in maternal plasma is an effective noninvasive prenatal testing technique that has been used in fetal aneuploidy screening worldwide. However, its clinical application is limited by the low fetal fraction (<4%) of cell-free DNA in many singleton pregnancies, which usually results in screen failures or no calls. In addition, dizygotic twin contributions of cell-free DNA into the maternal circulation can vary by 2-fold, complicating the quantitative diagnosis of fetal aneuploidy. We performed semiconductor sequencing of shorter fragments (107-145 bp) of circulating cell-free DNA to improve the fetal DNA fraction at lower uniquely mapped reads (1-8.5 MB) to reduce the probability of no calls. We identified 2903 plasma samples from pregnant women, including 86 dizygotic twin pregnancy, that were collected at a single prenatal diagnostic center between October 2015 and July 2018. Size-selection noninvasive prenatal testing for fetal aneuploidy was applied to 2817 plasma samples (1409 male and 1408 female fetuses) and 86 dizygotic twins using noninvasive prenatal testing with and without size selection. Shorter fragment size was the key factor affecting fetal fraction in multivariable linear regression models as well as to validate the accuracy of the size selection for noninvasive prenatal testing. Analysis of 1409 male fetuses by multivariable linear regression showed that maternal age, body mass index, number of pregnancies, average cell-free DNA size, maternal plasma cell-free DNA concentration, library concentration, and multiple gestation were negatively correlated with fetal fraction. Conversely, gestational age and uniquely mapped reads were positively correlated with fetal fraction. Compared with ≤120 bp cell-free DNA fragments, mean fetal fraction differences were -3.57% (95% confidence interval, -5.95% to -1.19%), for 121-130 bp, -9.52% (95% confidence interval, -11.89% to -7.14%) for 131-140 bp, and -14.47% (95% confidence interval, -18.37% to -10.58%) for ≥141 bp (Ptrend < .0001). These results were statistically significant after multivariable adjustments in models for fetal fraction. Meanwhile, results from 86 dizygotic twins showed that the size selection increased the fetal fraction by ∼3.2-fold, with 98.8% of samples reaching a fetal fraction >10%. Improved detection accuracy was also achieved. Sequencing shorter cell-free DNA fragments is a reasonable strategy to reduce the probability of no calls results because of low fetal fraction and should be recommended to pregnant subjects.