This archive contains the nanopore generated HCMV genomes from the urine and lung clinical samples. The illumina derived genome sequences have been uploaded to GenBank as they are of higher quality, the nanopore sequences are uploaded here to avoid duplication on GenBank. Raw FASTQ reads from the virus are uploaded to NCBI SRA. The full paper abstract is given below. Nanopore sequencing is becoming increasingly commonplace in clinical settings, particularly for diagnostics and outbreak investigations. Its portability, low cost and ability to operate in near real-time has propelled it to the forefront of the recent SARS-CoV-2 pandemic. Although high sequencing error rates initially hampered its wider implementation, improvements have continually been made with each iteration of the nanopore flow cells and base calling software. Here, we assess the feasibility of using nanopore sequencing to determine the complete genome of human cytomegalovirus (HCMV) present in high-titre clinical samples without viral DNA enrichment, PCR amplification or prior knowledge of the sequences. We utilised a hybrid bioinformatic approach that involved assembling the reads de novo, improving the de novo consensus through alignment of reads to the best-matching genome from a collated set of published genomes, and polishing the improved consensus. The final consensus genome sequences from a urine and a lung sample, the latter with an HCMV to human DNA load approximately 50 times lower than the former, achieved 99.97 and 99.93% identity, respectively, to the bench-mark consensuses obtained independently by Illumina sequencing. Thus, we demonstrate that nanopore sequencing is capable of determining HCMV genomes directly from high-titre clinical samples with high accuracy.