AudioCraft: Meta’s new AI tool for text-to-audio generation
Shemar-Leslie Louisy / Our Today
Meta has unveiled Audiocraft, an open-source, AI-powered tool that allows users to generate high-quality, realistic audio and music from text-based inputs.
The innovation, which was released on August 2, promises to revolutionise the creative industries such as music composition, game development, and social media content creation.
“We’ve seen immense progress in generative AI for images, video, and text in recent years, but audio generation has lagged behind due to its complexity. AudioCraft simplifies the process, giving people the tools to explore the vast potential of generative audio, be it for music, sound effects, or compression.”
Meta
The AudioCraft framework comprises three models: MusicGen, AudioGen, and EnCodec.
MusicGen generates music from text-based inputs and was trained on Meta-owned and specifically licensed music. AudioGen, on the other hand, generates environmental sounds and sound effects, such as a dog barking or footsteps on a wooden floor, from text-based descriptions.
AudioCraft’s capabilities extend to generating complex audio sequences, making it especially suitable for music composition. Its ability to capture long-range dependencies and stylistic nuances in music sets it apart from previous approaches that used MIDI or piano rolls.
Comments