Lyria 3 is Google DeepMind's newest and most advanced music generation model. It enables users to create original, 30-second audio clips with custom cover art from textual or visual prompts within the Gemini app. It builds upon previous audio generation by generating lyrics based on user prompts, allowing control over style, vocals and tempo to create realistic tracks. Lyria 3 is designed for original expression and creativity, not for mimicking existing artists.
Lyria 3 is available in the Gemini app to users aged 18 and over. Users can create music by describing a genre, mood, or memory, or by uploading a photo or video for inspiration. The model is capable of generating lyrics, and users have control over vocals, tempo and style. All generated tracks are embedded with SynthID, Google's imperceptible watermark for identifying Google AI-generated content. Google AI Plus, Pro, and Ultra subscribers will enjoy higher limits, though specific limits are not yet listed. The model is currently available in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese, with plans to expand language coverage. Pricing details were not found in the provided search results.
