Udio

You describe a song in text (genre, mood, optional lyrics) and Udio's model generates a complete 48kHz-stereo track - vocals and instrumentation together - in one pass, rather than layering instruments individually. Inpainting lets you regenerate just one section of an existing song in place, Extend continues a track past its current ending, and Audio-to-Audio remixing feeds an uploaded track back through the model so it can reimagine it, instead of every generation having to start from a blank prompt.