Lyria 3.5

Verified

Google DeepMind music model for full songs with vocals, lyrics and tempo control

Visit Lyria 3.5 →No rating yetFreemium

Product preview

Lyria 3.5

What it is

Lyria 3.5 is Google DeepMind's latest music generation model. From a text prompt, or even an uploaded image, it creates complete tracks of variable length up to about three minutes, with a beginning, development and ending. It can sing in several languages and styles, follow lyrics you write or generate lyrics around a theme, and cover genres from pop and funk to Motown and global styles. Prompts can specify tempo, key, vocal style and acoustic character for finer control. People can try it in Google Flow Music and the Gemini app, while developers can call it through the Gemini API with per-song pricing. Every track carries an inaudible SynthID watermark that identifies AI-generated audio.

Key features

Full songs up to about three minutes
Vocals and lyrics in several languages
Tempo, key and style controls
Image-to-music input

What we liked

+ Backed by Google DeepMind research
+ Available in consumer apps and via API
+ SynthID watermark for transparency

Worth noting

· API use is billed per song with no free API tier
· Access and limits vary by Google product

Editor's notes

More context from AI BirdView

Lyria 3.5 is notable for bringing a research-grade music model into everyday Google products. You can describe a song in a sentence, add your own lyrics, or even upload an image for inspiration, and get back a complete track rather than a short loop. Google says this version improves melody, lyric adherence and vocal quality over earlier releases.

Generates full tracks of variable length, up to about three minutes
Writes or follows lyrics and sings in several languages and styles
Prompts can set tempo, key, vocal style and acoustic feel
Reachable through Flow Music, the Gemini app and the Gemini API
Every output is marked with SynthID so AI-generated audio can be identified

Lyria 3.5 suits creators who want quick custom music for videos, podcasts or demos, as well as developers who want to add song generation to their own apps. Because Google publishes per-song API pricing, it is easy to estimate costs for a product feature. A practical tip: be specific about structure and mood in your prompt, for example naming the tempo, the genre and where the chorus should land, and keep your own lyrics short and rhythmic so the vocals follow them more closely.