On Wednesday, Google revealed that it’s introducing a music generation feature in the Gemini app, utilizing DeepMind’s Lyria 3 model, still in beta. Users describe the desired song to create, and the app produces a track with lyrics, such as a “comical R&B slow jam about a sock finding its match,” along with Nano Banana-designed cover art.
Users can upload a photo or video, and the AI generates a song matching the media’s mood. Lyria 3 surpasses previous models by creating realistic, complex tracks, allowing adjustments in style, vocals, and tempo. Google is also rolling out Lyria 3 to YouTube’s Dream Track, expanding its availability beyond U.S. YouTube creators.
While not designed for direct artist mimicking, naming an artist in a prompt inspires Gemini to create similar-styled tracks. A blog post clarifies that music generation with Lyria 3 aims for original expression, with filters checking outputs against existing content.
AI-generated songs with Lyria 3 will have a SynthID watermark, letting users verify AI generation. This feature, supporting multiple languages, is available globally to all 18+ Gemini users.
AI music gains varied reactions; while platforms adopt AI for monetization, music industry lawsuits address copyright concerns over training material. Tools are emerging to mark AI-generated music, reducing fraudulent streams.