Google Pushes Deeper into AI Music with Lyria 3

News

1. Deccan AI Raises $25 Million in Round Led by A91 Partners

2. Adani in Talks With Meta, Google, Flipkart on Data Centers: Report

3. India Ties $70 Billion Data Center Push to Local Build-out, Jobs

4. Novo Moves Delhi High Court Against Dr Reddy’s Over Olymviq Brand

5. Google Pushes Deeper into AI Music with Lyria 3

6. OpenAI Foundation to Deploy $1 Billion on AI Risks, Health and Jobs

Google has rolled out its latest music-generation models, Lyria 3 Clip and Lyria 3 Pro, to developers in public preview, expanding its push into generative audio through the Gemini API and a new music-focused workspace in Google AI Studio.

Google said the AI Studio experience requires a paid API key.

The new models are designed to generate structured music from text prompts and images, with support for vocals, timed lyrics, and song elements such as verses and choruses, Google said, adding that the models produce 48kHz stereo audio and can be used through standard Gemini API calls.

Google has launched the two variants aimed at different use cases. Lyria 3 Clip, listed in the API as lyria-3-clip-preview, generates 30-second clips intended for short-form uses such as previews, loops and social content.

While the company positions Lyria 3 Clip as a step forward in AI-assisted composition, early reactions suggest the technology is still evolving, particularly when it comes to replicating the nuance of human artistry.

At its core, it is designed to generate full musical compositions with a degree of structural coherence, handling elements like verses, choruses, and transitions in a way that feels continuous rather than fragmented. 

Developers can prompt the model using natural language to create songs with specific tempos, styles, or moods, and even align lyrics to precise timestamps within a track.

The other variant, Lyria 3 Pro, or lyria-3-pro-preview, is built for longer-form compositions and can create tracks of roughly three minutes with more developed song structure.

Google also highlighted multimodal features that let developers use images to shape the mood and atmosphere of a composition, alongside text-based controls for tempo and lyric timing.

It also showed the model being used to generate background music for uploaded videos and to build an alarm-clock app that creates personalized songs using details such as time, weather and calendar events.

Every track generated with Lyria 3 includes Google’s SynthID watermarking technology, which the company said is intended to help identify AI-generated audio even after it has been modified.

The rollout comes as Google broadens its AI music efforts beyond developer tools. 

TechCrunch reported that Lyria 3 Pro is also being added to Vertex AI in public preview, to Google Vids and to the Gemini app for paid subscribers, extending the model beyond standalone experimentation into enterprise and consumer products.

Early reactions on social media suggest that while the technology is advancing, it is not without limitations. Some users point to a gap between technical quality and emotional depth.

Others drew comparisons with competing tools such as Udio. 

“It is hard to explain, Udio has less dynamic range, but the vocals it produces are more human, more real, it can do a Celtic accent that will bring a tear to your grandfather’s eye in a way that is very hard to achieve with the “perfect” output of Lyria 3,” another user posted on X.

The rollout is also prompting new behaviours among creators. “New skill to master in 2026 – learning how to describe music with words,” another user said on X, pointing to the growing importance of prompt design in creative workflows.

At the same time, some see immediate practical value, particularly for content creation. “Cool Lyria 3 will transform entertainment industry,” yet another user said on X.