Last month, Google presented Lyria 3his latest generative model of music, capable of creating 30 second tracks to which is added a personalized cover made by Nano Banana. You can create themes even without the user providing the lyricssince this is generated from the user’s instructions. Lyria also allows you to control track elements such as style, voice and tempo.
Today, Google presented Lyria 3 Proits most advanced musical generation model. Can create tracks up to 30 minutesoffering users greater creative control and more customization options. For example, users can now enter instructions with details about items such as introductions, verses, choruses and bridges. As a result, Google claims the Lyria 3 Pro can create more realistic and musically complex tracks.
Google claims that ‘Lyria 3 Pro understands music composition better’ compared to the previous model and that it is ‘ideal for experimenting with different styles or generating songs with complex transitions’.
The company assures that the ‘responsibility was fundamental’ when designing and training this model, so only uses materials to which the company owns the rights.
He also points out that the model does not imitate any artist. However, it indicates that if users specify an artist at the prompt, the model is heavily inspired by that artist to generate a track.
As with AI-generated images and videos, all Lyria 3 and Lyria 3 Pro music creations come with a watermark. SynthID to identify content generated by Google AI.
It is now available for paying Gemini users and for business clients in Vertex AI. Additionally, developers have access to the tool through the API of Gemini and Google AI Studio. The company is also integrating it into Google Vidsan AI-based video generation platform.