Meta's MusicGen AI makes high-quality music clips from text — here's how to try it

Meta has released MusicGen, an open-source AI tool that generates new music by combining text prompts with existing audio clips. By analyzing the tonal and structural qualities of a provided melody, the AI can transform classical compositions into modern genres like synth-pop or hip-hop, demonstrating a significant advancement in AI-driven creative tools. This release highlights the growing accessibility of sophisticated generative models, allowing users to experiment with music creation without needing to build complex systems from scratch. However, the technology currently faces notable performance hurdles. Users encounter significant processing delays when generating short audio clips through web demos, and local execution demands high-end hardware with substantial GPU memory, limiting immediate adoption for average consumers. Furthermore, the tool relies heavily on input context; generating music from text alone yields mixed results, particularly regarding tempo and specific stylistic nuances, indicating that human guidance remains essential for high-quality outcomes. This announcement is relevant to open data because MusicGen is released under an open-source license, making its code and models publicly available for inspection, modification, and redistribution. This transparency fosters innovation within the community, allowing researchers and developers to study the underlying algorithms and potentially improve efficiency or integrate the technology into broader data ecosystems. By releasing these resources openly, Meta contributes to a more collaborative and auditable landscape for artificial intelligence development, setting a precedent for how proprietary tech giants can share foundational tools with the public.

Source: tomsguide.com
Published on 2023-06-15