- What: Google DeepMind launched Lyria 3 Pro, an advanced music generation model.
- Key Upgrade: Tracks now reach 3 minutes in length, a 600% increase over the standard Lyria 3’s 30-second limit.
- New Feature: "Structural Awareness" allows users to prompt for specific song elements like intros, verses, and choruses.
- Availability: Rolling out across Vertex AI, Google AI Studio, Gemini app (paid), Google Vids, and ProducerAI.
- Security: Every track includes SynthID, an imperceptible watermark for AI identification.
Google DeepMind has officially shattered the technical ceiling for AI-generated music with the release of Lyria 3 Pro, an advanced model capable of generating high-fidelity tracks up to three minutes long. This milestone, announced by Senior Product Manager Myriam Hamed Torres, moves AI music beyond short-form clips and into the realm of full-length compositions by introducing "structural awareness" that understands the nuanced architecture of music.
The launch signifies Google’s aggressive push to dominate the creative AI landscape, integrating Lyria 3 Pro across its entire ecosystem from enterprise-grade Vertex AI to the consumer-facing Gemini app. By allowing users to command specific transitions and song parts, Google is positioning Lyria 3 Pro not just as a generative toy, but as a professional-grade tool for musicians, developers, and marketing teams.
Structural Intelligence: The End of "Vibe-Only" AI Music
The most significant technical leap in Lyria 3 Pro is its ability to comprehend musical composition. While previous models often produced "musical wallpaper"—continuous loops without a clear beginning or end—Lyria 3 Pro allows for granular creative control. According to Google DeepMind, users can now prompt the model for specific elements such as intros, verses, choruses, and bridges.
This structural awareness ensures that a three-minute track maintains coherence and emotional arc. For developers and creators, this means the ability to generate songs with complex transitions and specific stylistic shifts. In testing, the model has demonstrated a deeper understanding of how different genres transition between high-energy choruses and stripped-back verses, solving one of the most persistent "hallucination" problems in audio AI: the loss of rhythm or key over longer durations.
Scaling Production from Vertex AI to Google Vids
Google is ensuring that Lyria 3 Pro is accessible at every level of the production stack. The rollout strategy covers four primary pillars:
1. Enterprise Scaling with Vertex AI Lyria 3 Pro is now in public preview on Vertex AI. This is designed for businesses that require high-fidelity audio on demand at a massive scale. According to the announcement, organizations can use the platform to generate bespoke soundtracks for gaming environments or integrate the model directly into their own creative and video platforms.
2. Developer Flexibility in AI Studio For developers building third-party applications, Lyria 3 Pro is available alongside Lyria RealTime in Google AI Studio and the Gemini API. This allows for the creation of "the next generation of creative tools" that leverage Lyria’s structural coherence to offer users more flexibility than a standard prompt-to-audio interface.
3. Video Creation in Google Vids In a move that targets the growing social media and corporate marketing sector, Google is integrating Lyria 3 and Lyria 3 Pro into Google Vids. The AI-powered video creation app will now allow users to generate custom music that matches the specific style and "vibe" of their projects. This feature is rolling out to Google Workspace customers and Google AI Pro and Ultra subscribers immediately.
4. Consumer Access via Gemini and ProducerAI Individual creators can access the longer 3-minute generations through the Gemini app, starting with paid subscribers. Furthermore, Google highlighted ProducerAI, a collaborative tool built by musicians. ProducerAI now offers an "agentic experience" using Lyria 3 Pro, helping artists iterate on comprehensive songs rather than just isolated loops.
The "Responsibility First" Framework
Acknowledging the sensitivities of the music industry, Google emphasized that Lyria 3 was developed in close partnership with the creative community. The model was trained on materials that Google and YouTube have the legal right to use under existing terms of service and partner agreements.
To protect human artists, Google has implemented two critical guardrails:
- Anti-Mimicry Filters: If a user includes a specific artist’s name in a prompt, the model is designed to take that as "broad inspiration" rather than mimicking the artist’s specific voice or style.
- SynthID Watermarking: All outputs from Lyria 3 and Lyria 3 Pro are embedded with SynthID. This technology places an imperceptible watermark into the audio, allowing it to be identified as AI-generated even if the file is compressed or modified.
Impact: Moving from Simulation to Collaboration
For the AI industry, the launch of Lyria 3 Pro marks a shift from audio simulation to genuine utility. For developers, the ability to generate structurally sound, long-form audio via API opens the door for dynamic soundtracks in video games that evolve with the player's journey. For marketers, it reduces the reliance on stock music libraries, allowing for hyper-personalized audio branding.
"This changes how developers will think about audio; we are moving from 30-second clips to full-scale musical architecture."
The integration into ProducerAI is perhaps the most significant for the industry. By framing the AI as a collaborative "agent" rather than a replacement, Google is attempting to bridge the gap between traditional music production and generative technology. Grammy-winning producer Yung Spielburg and legendary DJ François K have already been cited as early adopters, using the tool to score films and iterate on new releases.
What’s Next
The rollout of Lyria 3 Pro is currently focused on professional and paid tiers, including Google AI Pro and Ultra subscribers. As the model moves out of public preview on Vertex AI, the industry can expect a surge in third-party applications leveraging the API for specialized audio tools.
Google DeepMind has signaled that they will continue to expand the "Music AI Sandbox," inviting more artists to integrate these tools into their workflows. The focus remains on quality and coverage, with plans to expand the model's capabilities in more languages and regions over the coming months.