Possibility to generate huge amounts of RNA-Seq data created a demand for novel tools capable of analyzing large transcriptomic data sets. While reference-based transcriptome analysis prevails in medical studies, multiple research projects require de novo transcriptome assembly tools. Due to varying expression levels across different genes, RNA-Seq data sets are characterized by highly uneven coverage depth, which appears to be one of the challenges in the de novo transcriptome assembly. Since recently developed SPAdes assembler successfully addresses this problem we have decided to expand its functionality to enable high-quality transcriptome and metatranscriptome assemblies from short reads. We present rnaSPAdes — a new SPAdes-based RNA-Seq assembler, demonstrate its application on various datasets and compare with modern state-of-the-art transcriptome assembly tools. Also we show some algorithmic and biology challenges of metatranscriptome assembly and how rnaSPAdes copes with these problems and construct accurate reliable transcripts from different dissimilar organisms.