SUMMARY: Here, we present an approach, called SVUPP, which improves genotyping of structural variant (SV) by incorporating read phasing information into genotype likelihoods. Through comprehensive benchmarking, we show that SVUPP achieved higher accuracy than cuteSV2, Sniffles2 and kanpig with both long and ultra long Oxford Nanopore Technologies (ONT) data as well as Pacific Biosciences (PacBio) HiFi data for genotyping SVs without close neighbor SVs. SVUPP can be applied together with SV callers such as cuteSV2 and take the per-read phasing information from reference panel based phasing method such as QUILT2 or from reference-free phasing method such as WhatsHap. AVAILABILITY AND IMPLEMENTATION: SVUPP is written in Nextflow with modular design and is freely available here https://github.com/Zilong-Li/SVUPP.