Lyria 3 is presented as an advanced music generation model designed to facilitate the creation of high-fidelity audio tracks. Its core functionality revolves around generating music through user-provided prompts, emphasizing a "natural flow from note to note."

The model offers several key capabilities:

1. Image-to-Music Transformation: Users can upload an image and instruct Lyria to convert it into a custom, high-fidelity musical track, indicating a multimodal input capability where visual data influences audio output.
2. Detailed Compositional Control: It provides extensive control over musical parameters, allowing users to define specific elements such as "realistic vocal styles" and "acoustic preferences." This enables fine-tuning of the generated output to achieve a desired sound, suggesting a sophisticated understanding and manipulation of musical characteristics.
3. Professional-Grade Audio Export: Lyria 3 can produce "crisp, clear tracks" suitable for various professional projects, ranging from background ambience to mainstage anthems. This highlights its capacity to generate high-quality, production-ready audio.

Lyria 3 is part of a broader "Lyria model family," which collectively specializes in generating music. This family of models is capable of producing not only short clips and full tracks but also delivering a "constant stream of music," indicating versatility in output length and application. The model is accessible for use via Gemini, encouraging users to "express, explore, and experiment" with new genres and soundscapes. While specific technical methodologies are not detailed, the description implies an advanced artificial intelligence system capable of interpreting complex prompts and generating musically coherent and high-fidelity audio, possibly leveraging deep learning techniques for prompt-to-audio synthesis and multimodal integration.