The landscape of audio and music creation is rapidly evolving as major players in the tech industry, such as Meta and Google, introduce groundbreaking generative artificial intelligence (AI) software tailored for music and audio creation.
Meta, the parent company of Facebook, recently unveiled AudioCraft, a sophisticated multi-model generative AI tool. This tool employs artificial intelligence to produce high-quality audio and music content from text prompts. With AudioCraft, users can input text prompts to generate seamless and convincing audio and music compositions, eliminating the need for manual intervention.
The potential of AudioCraft is profound. Meta envisions scenarios where professional musicians can explore new compositions without playing any musical instruments and small business owners can effortlessly add soundtracks to their videos on platforms like Instagram. Meta’s AudioCraft encompasses three distinct AI models: MusicGen, AudioGen, and EnCodec. MusicGen focuses on generating music, utilizing Meta-owned and licensed music for training. AudioGen specializes in creating audio and was trained on public sound effects. The third model, EnCodec, concentrates on audio compression, enabling higher-quality music generation within smaller file sizes.
Mark Zuckerberg, CEO of Meta, announced that the code for AudioCraft will be open-sourced. Unlike solely pre-trained AI products, this open-source approach allows researchers and practitioners to utilize AudioCraft to further train models with their unique datasets, a critical feature for creative fields.
Meta emphasized the necessity for AudioCraft, as generative AI has seen significant progress in video, images, and text but has lagged in the audio domain. AudioCraft aims to bridge this gap by providing accessible and innovative audio generation tools that are user-friendly and open for further development.
In parallel, Google has joined the race with its AI offering, TextFX. Developed in collaboration with renowned hip-hop artist Lupe Fiasco, TextFX aids songwriters in crafting lyrics and themes by generating novel meanings, semantic connections, and exploration paths from user inputs. Lupe Fiasco, whose linguistic techniques contributed to the creation of TextFX, highlighted how the program facilitated the writing of a song in a mere two hours.
As AI technologies like AudioCraft and TextFX continue to shape creative domains, regulatory discussions ensue regarding AI’s distinctive aspects, especially copyright implications concerning the data used to train AI models. The EU has taken a lead in this conversation, introducing a draft law to address these evolving challenges.
In the ever-evolving landscape of creative AI tools, Meta and Google are at the forefront, ushering in a new era of accessible and innovative audio and music generation technologies.