Music generation

AI music generation covers a spectrum from simple background music loops to complete songs with vocals, lyrics, and production. The leading tools, Suno and Udio, sit at the full-song end: you describe genre, mood, tempo, and lyrical theme, and get back a finished track in under a minute. ElevenLabs launched their own AI music generator in August 2025, positioned for commercial use and trained on licensed data. Google's Lyria model powers music features inside Google products.

The underlying technology is typically a combination of large language models for lyric and structure generation and audio diffusion or transformer models for actual sound synthesis. The models have been trained on vast libraries of music (a source of ongoing copyright litigation with major labels), and they've internalized an enormous range of genres, production styles, and musical structures.

For builders, AI music generation is most relevant in three contexts: content creation workflows (background music for videos, podcasts, ads), games or apps that need dynamic adaptive soundtracks, and personal creative tools. The commercial licensing situation matters here: Suno's paid plans come with commercial rights, but the provenance of training data remains legally contested. ElevenLabs' music offering being trained on licensed data is a meaningful differentiator for businesses that need clean IP. The field is also moving fast, with capabilities for stem separation (splitting a track into individual instrument parts), MIDI export, and style control all landing in 2025.