Key takeaways
- Meta has open sourced AudioCraft, a generative AI system for creating high-quality audio and music from text.
- AudioCraft features three core models: MusicGen, AudioGen, and EnCodec, providing a unified solution for audio generation.
- The latest version enhances the previous MusicGen, offering improved sound quality and reduced glitches.
- AudioCraft simplifies the development of sound generators and compression algorithms within a single code base.
- Meta anticipates that AudioCraft will significantly impact future technologies and the way users engage with audio.
Meta today announced that it is open sourcing AudioCraft, a new family of generative AI models built for generating high-quality, realistic audio & music from text.
AudioCraft introduces a unified code base that encompasses music, sound, compression, and generation functionalities, providing a comprehensive solution in one place. The system comprises three models: MusicGen, AudioGen, and EnCodec.
The new release of AudioCraft is an improvement over the previous MusicGen version. It includes a better EnCodec decoder that allows for higher-quality music generation with fewer glitches. Moreover, it now has pre-trained AudioGen models, enabling the system to create environmental sounds and sound effects like a dog barking, cars honking, or footsteps on a wooden floor.
This release is exciting as it simplifies building on the top of the state of the art in audiogeneration. People can now build things like sound generators and compression algorithms with the same code base.
Meta in its blog stated that AudioCraft represents a significant advancement in generative AI research. They believe that the straightforward approach they developed for audio generation will have a profound influence on future technologies and how we interact with them.
Meta expressed its excitement over the creative potential of people using AudioCraft and looks forward to seeing what they will create with it.
Access the code here: https://bit.ly/3QnMya3
