Smartphone-based palmprint biometrics has received limited attention from the researchers. This work presents an effective framework for generation of synthetic data for smartphone-based palmprint biometrics. A new smartphone-based video palmprint dataset is collected with six notable variations in each hand video. The dataset is acquired from 30 subjects with 60 classes i.e. 2 classes per subject covering the left and right hands. The synthetic images generated using the collected smartphone-based palmprint data are observed to be modeling the underlying variations effectively. The quantitative evaluation is furnished through Fréchet Inception Distance (FID) and Learned Perceptual Image Patch Similarity (LPIPS), meant to measure the realistic appearance of the synthetic images. The results demonstrated good synthetic palmprint images with an FID score of 150–170 and LPIPS score of 0.524–0.605, which reflect the efficacy of the proposed framework.