Cell and tissue functions arise from complex interactions among numerous genes, and a systematic understanding of these functions requires isoform-resolved, whole-transcriptome-scale analysis of single cells with high spatial resolution. Here, we introduce a spatial transcriptomics method based on a novel in situ RNA amplification strategy, enabling short RNA sequence detection and hence spatially resolved expression profiling of individual cells at the whole-transcriptome scale with splice-isoform resolution. Using this approach, we imaged ~33,000 distinct RNAs, including ~23,000 genes and ~10,000 isoform-defining transcripts, in the mouse brain. Our data enabled systematic analyses of region- and cell-type-specific gene programs and ligand-receptor-based cell-cell communications. These data further revealed a rich spatial diversity and cell-type specificity in isoform usage across numerous genes and identified brain structures particularly enriched for specific isoform usage. We anticipate broad application of this method for molecular and cellular analysis of tissues, unlocking previously inaccessible discoveries in cell and organismal biology.