← All models

Lyria 3.5

Published Publisher Googlelyria-3.5

Overview

Lyria 3.5 is Google’s music-generation model, optimized for full songs with coherent verses, choruses and bridges. It uses latent diffusion applied to audio representations. (aistudio.google.com)

Release and availability

Google announced Lyria 3.5 on July 29, 2026, rolling it out in Google Flow Music. On September 4, 2026, Google announced availability in the Gemini app and Gemini API, as well as Google AI Studio and Google Vids. (blog.google)

Capabilities and uses

  • Accepts text and image inputs and produces MP3 audio and text lyrics. The documented input limit is 131,072 tokens. (aistudio.google.com)
  • Generates 44.1 kHz stereo music, including vocals, lyrics and instrumental arrangements. Full songs generally last a couple of minutes; prompts can influence duration, tempo and structure. Google highlights improvements in musicality, lyric adherence, vocal expression and pronunciation. (aistudio.google.com)
  • Potential uses include video backing tracks, brand jingles, personalized ringtones and other music creation. (blog.google)

Limitations

  • Gemini API music generation is single-turn; multi-turn refinement of a generated clip is not supported. The same prompt can produce different results. (aistudio.google.com)
  • Safety filters can block requests, including those for specific artist voices or copyrighted lyrics. Generated audio carries a SynthID watermark. (aistudio.google.com)
  • The model does not support context caching, Batch API, Live API, function calling or search grounding. (aistudio.google.com)

Pricing

For the Gemini Developer API paid tier, Google lists USD $0.08 per full song; a free tier is not available for this model. The published rate is per song, not separate input- and output-token rates; separate input and output charges were not verified. Pricing checked November 16, 2026. Other products or platforms may have different terms. (aistudio.google.com)

Official sources