Google published a blog post yesterday (July 29), announcing the rollout of the Lyria 3.5 music generation AI model in Google Flow Music. The model improves musicality, lyrics, vocal quality, and more, enabling users to create richer musical works.
According to Google, compared with the previous-generation model, Lyria 3.5 focuses on improving melodic arrangement, lyric generation, vocal performance, and song control. The new model not only supports generating richer and more complex melodic structures, but also better follows users' prompts, understands song intent, and improves the quality of generated songs.
As for vocals, Google says that those generated by the new model sound more like real singing, with more nuanced emotional expression and more accurate pronunciation. In addition to content generation quality, Lyria 3.5 also enhances creative control options, allowing users to set the song's tempo and duration more flexibly.
Google also demonstrated the song “Hasta Que Salga El Sol,” created using the Lyria 3.5 model. The related video is below: