Latest models
- Lyria 3.5 Current
Google's music model for full-length songs with verses, choruses and bridges, producing 44.1 kHz stereo audio from text or up to ten images through the Gemini API. Source: Gemini API docs ↗
Checked October 7, 2026. See the full AI model tracker.
What we like
- Songs with verses, choruses and bridges, about a couple of minutes long
- 44.1 kHz stereo audio, MP3 by default with WAV available
- Accepts text or up to ten images as input
Watch out for
- Mostly aimed at developers via the API
- Check Google's terms for commercial use
- Duration and structure are steered by your prompt
Lyria is Google DeepMind’s music generation family. Lyria 3.5 is Google’s flagship music model, available through the Gemini API.
What is current
Google’s docs describe Lyria 3.5 as generating full-length songs with complex structure, including multiple verses, choruses and bridges, that last a couple of minutes, controllable through the prompt. It outputs 44.1 kHz stereo audio, MP3 by default and WAV on request, and accepts text or up to ten images. You can use it through generateContent or the Interactions API. See the model tracker.
How prompting works
State the genre, instruments, vocals, mood and structure. You can influence length by saying how long the song should be, or define the structure with timestamps.
Who should use it
Developers and creators who want structured songs from text or images. For a consumer music app see Suno.