Nanopore sequencing technologies applied to transcriptome analysis suffer from high error rates, limiting them largely to reference-based analyses. Here, the authors develop a computational error correction method for transcriptome analysis that reduces the median error rate from ~7% to ~1%.