Lyria 3.5 Release Notes / What's New

Google DeepMind has launched Lyria 3.5, a music generation model integrated into Google Flow Music. The update focuses on enhancing the overall quality of generated tracks through improvements in melodic complexity, lyrical structure, and vocal realism.

Enhanced Musicality and Melodic Structure

Lyria 3.5 enables the creation of richer and more complex melodic structures. These advancements are designed to make generated music sound more natural compared to previous iterations.

Improvements to Lyric Generation

Lyric generation in Lyria 3.5 has been upgraded to produce higher quality text with a specific focus on two areas:

  • Prompt Adherence: The model more accurately follows the user's specific instructions for the lyrics.
  • Structural Awareness: The model demonstrates a better understanding of song structure, leading to more coherent lyrical arrangements.

Vocal Quality and Expression

Lyria 3.5 introduces more realistic and emotionally nuanced vocals. These improvements include:

  • Emotional Expression: Vocals now carry more expression and emotion to enhance the song's impact.
  • Pronunciation: The model features improved pronunciation for clearer vocal delivery.

Creative Control for Users

Google Flow Music now provides users with more precise creative control over the technical aspects of their outputs. Specifically, users can more easily manage and adjust the tempo and duration of the tracks they generate.

Sources