New Delhi: Tech giant META announced that it will open its AI source code ‘AudioCraft’ to allow users to generate high-quality audio and music from text. The latest tool has the potential to revolutionize the music and audio industry by creating new compositions without the need to play a single note.
By open-sourcing this tool, META is enabling researchers and practitioners to access it and train their own models with their datasets for the first time, thus advancing the field of AI-generated audio and music. According to the META blog, “AudioCraft works for music, sound, compression, and generation — all in the same place. It is easy to build on and reuse, enabling people who want to develop better sound generators, compression algorithms, or music generators to do so in the same code base and build on the work of others.”
AudioCraft comprises of three models:
MusicGen – This model generates music from text prompts and is trained using META-owned and specifically licensed music.
AudioGen – It generates audio from text prompts and is trained on public sound effects.
EnCodec – This model allows for higher-quality music generation with fewer artifacts and is an improved version of the EnCodec decoder.
In addition to this, META has also released pre-trained AudioGen models that allow users to generate environmental sounds and sound effects like a dog barking, cars honking, or footsteps on a wooden floor.
“The AudioCraft family of models is capable of producing high-quality audio with long-term consistency and is user-friendly. AudioCraft simplifies the overall design of generative models for audio compared to prior work in the field. It provides users with the full recipe to work with the existing models that META has been developing over the past several years while also empowering them to push the limits and develop their own models,” META blog added.
Click on this link to listen the AI-generated music and audio.
Will it be a threat for musicians and sound designers?
Regarding concerns about the potential threat to musicians and sound designers, the new tools have the power to generate music and audio without the need for any instruments or musical expertise. While they might have the potential to replace certain jobs in the industry, it is still early to draw concrete conclusions, similar to the impact of ChatGPT on content writers.
However, META is promoting these models as helpful tools for industry players and believes they can enhance creativity. They see the AudioCraft family of models as tools for musicians and sound designers to provide inspiration, help people quickly brainstorm and iterate on their compositions in new ways.