Music via prompt: Google's Lyria 3 Pro composes complete songs
Google's new model Lyria 3 Pro can generate music pieces up to three minutes long and is integrated into several AI services.
Google introduces a new AI model for music creation, designed to generate more complex pieces.
(Image: Billion Photos / shutterstock.com)
With the new Pro version, Google is significantly expanding its Lyria‑3 model, which was already introduced in February. Previously, the music AI could only generate very short pieces with up to 30 seconds of playback time. As Google reports, the focus is now on longer compositions with a comprehensible structure.
Broader integration into Google's products
Users can specify elements such as intro, verse, chorus, or bridge, as the system is expected to recognize musical structures better than the previous version. This also increases control over the style and progression of a generated piece of music.
The new model will be integrated into several services. Lyria 3 Pro is available, among other places, in the Gemini app for paying subscribers, in Google Vids for video creation, and on Vertex AI for corporate customers. Developers will also gain access via the Gemini API and Google AI Studio. The music tool ProducerAI, recently integrated into Google Labs, also uses the system for song production.
Cooperation with the music industry and legal questions
Google emphasizes that Lyria 3 Pro does not copy specific artists. If a name is mentioned in the prompt, it serves merely as a rough stylistic guide. Additionally, Google automatically checks the generated content for possible matches with existing music.
Videos by heise
All generated tracks will be provided with the invisible SynthID watermark to identify them as machine-generated. According to Google, it uses content for training for which usage rights exist through YouTube, partner contracts, or applicable law. The company does not provide details about the datasets.
While Google positions its systems as support for creative processes and collaborates with producers such as Yung Spielburg or François K, generative AI remains controversial in the music industry. Other providers, like Suno, face lawsuits from music labels. The Society for Musical Performing and Mechanical Rights (GEMA) already demanded license fees from Suno at the beginning of 2025.
(joe)
