RNA splicing is a key mechanism linking genetic variation and complex disease, including in neuropsychiatric diseases such as schizophrenia. Splicing profiles are particularly diverse in the brain, but it is difficult to accurately identify and quantify full-length isoforms using standard “short-read” sequencing. CACNA1C is a large gene that shows robust genetic associations with many psychiatric disorders and encodes multiple, functionally-distinct voltage-gated calcium channels via alternative splicing. However, due to its size and complex splicing profile, the splice isoforms that are expressed in brain and associate with disease risk remain poorly understood. Using long-read nanopore sequencing, we characterised the expression and splicing profile of CACNA1C in human brain. We show that its splicing profile varies between brain regions and is substantially more complex than previously appreciated: we identified 38 novel exons and 83 high confidence novel isoforms, many of which are predicted to alter protein function. Surprisingly, we find little evidence for the existence of most “known” isoforms. Applying the same methodology to other neuropsychiatric disease genes gave similar results, indicating that the splicing profiles of many genes remain poorly understood. Our findings showcase the significant potential of long-read sequencing to characterise human splice isoform diversity and to facilitate the identification of disease-linked isoforms.