The new AI Lyria 3 represents Google DeepMind’s latest milestone in multimodal sound generation, now integrated directly into Gemini. The model allows users to transform text descriptions, images, or short videos into high-fidelity original music tracks. As reported in the official Google DeepMind release, the rollout began on February 18, 2026, for users aged 18 and over.
Key Takeaways
- Generation of 30-second tracks via text prompts, images, or video.
- Multilingual support (including Italian) and automated lyric writing.
- Copyright protection via SynthID invisible watermarking.
How the new AI Lyria 3 works on Gemini
The integration of the new AI Lyria 3 within the Gemini ecosystem drastically simplifies the composition process. Users can request tracks by specifying genre, tempo, and emotional mood. The system does not just produce melodies; it is capable of generating realistic vocal performances and professional arrangements.
Multimodal creativity and automated lyrics
In addition to instrumental music, the model supports automated lyric writing, synchronizing them with the produced rhythmic base. The versatility of the new AI Lyria 3 extends to its ability to interpret visual inputs: by uploading an image or video, the AI can compose a soundtrack consistent with the atmosphere and content of the analyzed scene.
Large Language Models (LLM): quando l’intelligenza artificiale impara a parlare
