Google officially launched a new music-generation feature for the Gemini app this Wednesday, leveraging the advanced DeepMind Lyria 3 model. Currently in beta, this tool allows users to transform text prompts, photos, or videos into original, 30-second musical tracks complete with lyrics and custom cover art.
How Gemini’s Music Generation Works
Users can generate audio by simply describing their desired sound. For instance, requesting a “comical R&B slow jam about a sock finding its match” triggers the AI to produce a full track. Beyond text, the tool is capable of analyzing uploaded photos or videos to compose music that aligns with the visual mood of the media.
Advanced Control and Global Expansion
The Lyria 3 model represents a significant leap in performance, delivering higher complexity and sonic realism compared to its predecessors. Users gain granular control over the output, adjusting elements such as tempo, vocal style, and musical genre.
Simultaneously, Google is expanding its “Dream Track” feature globally. Previously restricted to U.S.-based YouTube creators, this tool now provides international creators with the ability to integrate AI-generated music into their content.
Ethical Guardrails and SynthID Watermarking
Google has implemented specific constraints to address copyright and artistic identity concerns. While users can reference an artist’s name in a prompt, the system does not mimic specific singers. Instead, it uses the input as “broad creative inspiration” to match a mood or style. The company also employs internal filters to check generated outputs against existing copyrighted content.
To ensure transparency, all audio generated via Lyria 3 is embedded with a SynthID watermark. This technology serves as an identifier for AI-produced content, and Gemini now allows users to upload tracks to verify if they were created using AI.
Availability and Industry Context
The feature is currently rolling out to all Gemini users aged 18 and older globally. Supported languages include English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. This release arrives as the music industry continues to navigate a complex landscape of AI integration, facing ongoing legal scrutiny regarding training data copyrights while major platforms simultaneously move to monetize AI-generated compositions.
For more details on the technology, read the official Google blog post.