Meta, the technology giant, has unveiled their latest innovation in the field of artificial intelligence (AI): AudioCraft. This groundbreaking open-source code allows users to create music and sounds using generative AI.
AudioCraft comprises three AI models, each responsible for different aspects of sound generation. The first, called MusicGen, is capable of generating music based on text inputs. To train this model, Meta utilized an impressive 20,000 hours of music owned by the company or specifically licensed for this purpose.
Meanwhile, the second model, known as AudioGen, is designed to create audio from written prompts. This AI simulation is capable of replicating sounds like barking dogs or footsteps. To train this model, Meta used publicly available sound effects.
Meta recently granted the media access to samples of audio created with AudioCraft. The samples showcased an impressive level of realism, including natural whistling, sirens, and humming. However, some elements, such as guitar strings, still felt somewhat artificial.
The integration of AI into the music industry is not entirely new. Google, for instance, has already explored the fusion of music and AI with its MusicLM. AI-generated songs have garnered substantial attention and support, with renowned musicians like Grimes encouraging people to incorporate their voices into AI-made compositions.
While Meta recognizes that AudioCraft could initially find its niche in elevator music or stock songs, the company firmly believes that its potential extends far beyond these realms. In fact, Meta envisions AudioCraft ushering in a new wave of music, akin to the transformative impact that synthesizers once had.
Developing AI models for music is not without its challenges. The complexities of audio compared to written text models present unique difficulties. Meta acknowledges this and underscores the importance of open sourcing to ensure a diverse range of data is used to train AudioCraft, thereby reducing potential biases.
Nevertheless, not all segments of the music industry have welcomed AI with open arms. Concerns have been raised by record labels and artists regarding the use of copyrighted materials for training AI models and the subsequent potential legal issues that may arise.
As Meta introduces AudioCraft to the world, it remains to be seen how this powerful AI tool will shape the future of music creation. With its vast potential and Meta’s commitment to diversity and openness, it is certain to make waves in the industry.
