This post discusses DiffRhythm, an innovative technology capable of generating full-length songs quickly using latent diffusion models. The comments express excitement about the technology's speed but highlight shortcomings in song structure, noting that the outputs lack recognizable choruses and exhibit noticeable artifacts upon repetition. Many commenters believe improvements are needed for practical use in music creation, such as better tools for multi-tracking and generation of counterpoints. Despite mixed reviews about its current state, the potential to assist amateur music writers is seen as a significant opportunity.